A program for cell motility assessment with a batch process function?

A program for cell motility assessment with a batch process function?

We are searching data for your request:

Forums and discussions:
Manuals and reference books:
Data from registers:
Wait the end of the search in all databases.
Upon completion, a link will appear to access the found materials.

Cell motility assessment is a branch of experimental biology or medical science. One example could be an assessment of treatment effects on sperm motility of an animal. The standard procedure involves taking film clips of sperm (or any other moving objects) through a microscope and measuring average linear velocity of particles on these film clips. The program used for measuring these linear velocities is important, and there are not too many choices to my knowledge. However, my knowledge might be limited.

I have previously used CellTrak for measuring sperm swimming speeds. The program works well, but lacks a batch processing function. As the assessment bases on a considerable number of replicates, I end up with 1000-2000 film clips to be analyzed and clicking my way through CellTrak is a tedious and irritating process. Also CellTrak licence costs a lot of money for such a poorly written program.

Therefore, my question is: Is someone aware of a similar program to CellTrak, which includes batch processing option, or even better, which works through command line and can be looped? Are there open source options or extensions to widely used open source programs (ImageJ, Bioconductor, etc.)? Please give a short tutorial/personal experience to the usage of the suggested program (if any).

ImageJ has several tracking plugins, a good one being TrackMate. Most of it's functions can be scripted in various languages and the Fiji distribution can also run in headless mode. It's open source and won't cost you any mony for a much lager feture set.

I personally have used ImageJ in headless mode scripted with its own macro language because it is relatively easy to learn. TrackMate also should output average speeds for the tracks, if I remember correctly. Otherwise you would have to write a script to measure those from the tracking data. You can use the IJ macro language, Python, Java or one of the other supported languages.

In combination with a bash script wrapper for its headless mode I even use it in Makefiles to convert my tiff stacks to avi and put scale bars and time stamps on everything.

Hope this helps.

A benchmark of batch-effect correction methods for single-cell RNA sequencing data

Large-scale single-cell transcriptomic datasets generated using different technologies contain batch-specific systematic variations that present a challenge to batch-effect removal and data integration. With continued growth expected in scRNA-seq data, achieving effective batch integration with available computational resources is crucial. Here, we perform an in-depth benchmark study on available batch correction methods to determine the most suitable method for batch-effect removal.


We compare 14 methods in terms of computational runtime, the ability to handle large datasets, and batch-effect correction efficacy while preserving cell type purity. Five scenarios are designed for the study: identical cell types with different technologies, non-identical cell types, multiple batches, big data, and simulated data. Performance is evaluated using four benchmarking metrics including kBET, LISI, ASW, and ARI. We also investigate the use of batch-corrected data to study differential gene expression.


Based on our results, Harmony, LIGER, and Seurat 3 are the recommended methods for batch integration. Due to its significantly shorter runtime, Harmony is recommended as the first method to try, with the other methods as viable alternatives.


Cell migration plays an essential role in many biological processes: e.g. wound healing, embryogenesis, inflammation, and metastasis where uncontrolled cell migration can lead to tumor spreading and hence can cause cancer progression. Studying these processes frequently require cell tracking, and most motility studies of monolayer cultures involve fluorescent labeling of cells, which allows for fluorescence microscopy studies [1][2]. This technique either requires extensive mutagenesis to have fluorescent protein expressed by the cell type under study, or is limited by the fact that membrane-attached or permeable fluorescent drugs often alter cell behavior [3]. In sparse cultures, cells can have sufficiently good contrast against the background that their boundaries can be identified with bright field microscopy without labeling [4a][4b]. This can be done manually with point-and-click methods at great expense of time and labor [5][6][7][8][9][10]. When cells in only a few images have to be tracked, this is not a challenge. However, when long time-lapse sequences of motile cells need to be analyzed, this approach is impractical, raising the need for a reliable automated cell tracking program and assessment of the robustness of the method of analysis.

Representative programs for sparse cell culture analysis are listed in Table 1 . The majority of such programs (not included in Table 1 ) are designed for tracking fluorescently labeled cells [1][2] [15][16][17][18]. The programs designed for use with light microscopy images track either the cell nucleus [19][20] or the entire cell [21][11][22][12]. The cell position can be defined as: (i) the center of the nucleus [6] (ii) the centroid of the cell's perimeter as seen in the light microscope [23] (iii) the centroid of the cell's footprint as seen in the light microscope [4] and (iv) the centroid of the actin cytoskeleton of fluorescently labeled cells [24]. Most of these programs are not open source, and may be difficult to adapt to the specific purposes of a given experiment. In other cases the complexity of the mathematical procedures used for boundary identification may be a hurdle to adapting the code to a specific purpose [2][20][17]).

Table 1

Overview of representative programs for cell tracking by time-lapse light microscopy.

ProgramPublisher/SellerCommercialOpen sourceCoordinate output format
Autozell [11] Universität BremenYesNoText
TLA [12] University of UlmNoYesExcel/Text
CellTrack [13] Ohio State UniversityNoYesText
ImageJ plugins [14] ETH & UCSFNoYesText

There are two main approaches to cell tracking in the current state-of-the-art [25][21][16][20][26][18]. One approach is frame-by-frame image segmentation and tracking [15][27]. In the first step, the object candidates are detected in a given frame on the basis of their specific properties (border, texture, color). This approach is efficient when object borders are sharp, and it is commonly used with fluorescently labeled cells and other high-contrast images. The other approach consists in optimizing a parameterized model shape to fit the model to the cells in a frame. Instead of tracking all objects in the frame, this method focuses on those candidates which correspond to the chosen model shape [28][29]. As with the first approach, detected objects are paired between consecutive frames in order to produce tracks.

In order to address the issue of background removal, and to compare the performance of different programs on real data against a baseline which we fully understand and control, we here introduce a new cell tracking program, which we call PACT (rogram for Āutomated ell racking). PACT is suitable for tracking motile cells on flat and nanostructured surfaces, and is simple enough for users to freely modify it according to their experimental needs. Since nanostructured surfaces are currently of great interest as cell culture substrates and can appear as a highly non-uniform background in the image, we emphasize the use of a reliable spatial image filter here – see details in the materials and methods section, and file for the Matlab code.

Test results are presented for the performance of PACT, which is also compared to the performances of other programs (TLA [12] and Autozell [11]): efficiency with regard to object detection, accuracy of centroid positioning, and segmentation performance in the context of time-lapse analysis. A statistical analysis is then performed on the ensemble of individual tracks of cells to determine the overall cell population motility statistics in a movie of NIH 3T3 fibroblasts on glass. From the tracks, the auto-covariance function of the velocity is estimated (Equation S1 in supp. info.). It is well described by a simple exponential function with characteristic time, P, the persistence time of the motility (data shown in supp. info sect. 4). We also determine the amplitude of the velocity auto-covariance function, φ0, which is approximately equal to the mean squared velocity of the cells. Results of this analysis obtained with PACT, TLA and Autozell are compared.

Despite the need for cell tracking programs and algorithms, we found few studies evaluating their performance [11][12][26]. To our knowledge, this paper is the first comparative analysis of tracking programs with the scope of providing reliable data acquisition routines for developing motility models. We find considerable sensitivity of results to the specific algorithm/software employed. Thus, the tracking algorithm and its effect on results should be well documented in future studies, for instance as described in this study.

2. Implementation

The overall approach for integrative analysis of motility by TIAM is summarized in Fig.ਁ . Detection, tracking, feature extraction, and track editing algorithms were implemented in MATLAB (from MathWorks). The user interface to facilitate user-inputs was implemented in Java. A second user interface for dynamic visualization of individual or pairs of tracks was implemented in MATLAB. The TIAM software project has been deposited in GitHub for free access to the source code ( A detailed user guide, demo and the URL link for benchmark datasets are provided in the Github repository. Additional description of algorithms can be found in the Supplementary methods section.

Overview of the schema for data integration in TIAM. Transmitted light images are used for detecting and tracking cells. Several parameters quantifying the motility characteristics are calculated and stored in MATLAB  �ll arrays’. Individual tracks are considered for extracting information from reflection and fluorescence images that are part of multi-channel time-lapse data. Centroids from track positions are used for local segmentation and outlining that would correspond to the cell under consideration. Features are computed from the outlined regions and stored along with rest of the track-related information.

2.1. Detection of cells

TIAM is equipped to detect and track cells in transmitted light image series, such as those acquired by bright-field, differential interference contrast (DIC), or phase-contrast microscopy. We chose this approach for multiple reasons: a) Cell boundaries can be difficult to discern from fluorescence information when cells are in a crowded environment the inherent nature of transmitted light imaging ensures that cell boundaries provide some contrast even in a crowded environment. b) Using transmitted light imaging for tracking of cells frees up a fluorescence channel for acquiring additional information about cells' behavior. c) Using transmitted light microscopy instead of fluorescence microscopy allows for long-term live-cell imaging as phototoxicity is minimized.

TIAM's cell detection strategy involves finding cell-shaped patterns in the set of edges detected in an image. A Canny edge filter (Canny, 1986) is used to produce a binary image depicting all edges in a given video frame, and a circular Hough transform (CHT) (Duda and Hart, 1972) operates on this binary image to detect individual cells ( Fig.ਂ a𠄽). This two-step strategy has been applied previously to detect nuclei in zebra fish embryos (Melani et al., 2007). The Hough transform is a robust method for detecting parameterized curves in images, where the task of detecting complex patterns of pixels (a costly global search problem) is transformed into the task of constructing peaks in a parameter space. The Hough transform carries out a voting process, where each edge pixel casts votes on curve parameters with which it is consistent afterwards, the locations in the parameter space that have gained a sufficient number of votes are returned. Local maxima in this parameter space can be thought of as centroids of cells. This strategy is beneficial for detecting cells with low-contrast boundaries due to the ability of the CHT to detect shapes based on non-contiguous and partial set of edges. Furthermore, it bypasses the need for segmentation of individual cells and thus aid in the accuracy of detection in high-density environments (Fig. S1 for example). We have used Tao Peng's implementation of the CHT (CircularHough_Grd from the MATLAB File Exchange repository) as it considers a radius range during the voting process and includes an additional parameter for searching maxima over imperfect circular shapes. Accordingly, we have found our implementation to detect polarized T cells as well as cells of different types, morphologies and at different cellular densities in images acquired by all three aforementioned transmitted light microscopy techniques ( Fig.ਂ , Fig. S1, Fig. S2, and Videos S1𠄳). The individual parameters involved in the detection step are described further in the Supplementary methods section. Parameter values typically used in our T cell imaging experiments are also provided.

Detection and tracking of cells by TIAM. TIAM uses transmitted light images for detecting and tracking cells. Illustration of detection by TIAM is provided with an example (a𠄽). A DIC image of human primary CD8 T cells is used (a). The panels, b to d, represent sequential stages during cell detection. In the first step, the Canny edge filter is applied to generate a binary image of cell boundaries (b). Then, a circular Hough transform (CHT) is applied to this binary image. This operation maps cell outlines to points in a parameter space based on a voting scheme (c). Local maxima in the parameter space are used to pick centroids of cells (d). TIAM has a graphical user interface that walks the user through the choice of parameters for edge filtering and CHT to allow for accurate detection of cells.

Successful detection is critical for all the ensuing computational steps. Therefore we have developed a graphic user interface in Java to interactively change parameters of the Canny-edge filter and CHT to achieve successful detection of cells in transmitted light images. The user guide provides an example of this process to help with intuitive selection of parameter values. The user is prompted to adjust the scale of the image such that the cell size is similar to the example provided in the user guide. This attempts to ensure that the default radius range used during CHT voting process works well. Similarly, edge detection and additional CHT parameters can be chosen by comparison to the example images of these stages. The centroid positions are transformed back to the original scale at the end of the detection step, before proceeding with tracking cells.

2.2. Tracking

Tracking in TIAM is carried out in two steps. In the first step, a modified nearest neighbor association algorithm is applied to the outputs of the cell detection step to yield short track ‘segments’ (Fig. S3a). At each time step t, each cell is linked to the spatially nearest detected cell of the previous time step t −ਁ, provided the nearest detected cell is within a maximal allowed distance r. This process proceeds in this manner only when cells are sufficiently separated and there is no tracking ambiguity. If there is more than one cell within r, the algorithm returns the track segment that has been produced up to that frame and initiates new tracks with neighboring cells that caused the ambiguity. This typically happens in cases when cells cross paths or where they are present at high local density. Thus, these track segments represent sequences over which the algorithm can confidently provide tracking results. We preferred the nearest neighbor algorithm for its simplicity and intuitiveness, both in implementation and performance, when compared to the state of the art model-based tracking approaches. In addition, we prefer to use longer time-intervals to reduce phototoxicity during long-term (over an hour) multi-channel time-lapse imaging. With T cells being highly motile, longer time-intervals may not provide overlapping cells in subsequent frames, which is a restrictive requirement of contour evolution based techniques (Padfield et al., 2011). Although the nearest neighbor algorithm fails to perform well at high cell densities, as discussed later, we have obtained accurate tracking with about fifty cells in the field of view.

In the second step, an assignment algorithm is used to join shorter segments end-to-end into longer cell tracks (Fig. S3b). In order to perform segment joining, a similarity is first defined between every pair of segments based on compatibility factors such as their start/end frame, location, and speed. Then the Hungarian algorithm (Munkres, 1957) is used to find a globally optimal mapping between segments based on the similarity matrix (Bise et al., 2011 Jaqaman et al., 2008 Perera et al., 2006). Out of these mapped assignments, segments are only joined if their similarity falls above some threshold. The two-tiered approach to tracking aims to be computationally efficient by implementing an unsophisticated, greedy nearest neighbor algorithm when the tracking scenario is simple, and a more complex set of computations using the nearest neighbor results when the tracking scenario is ambiguous.

The tracking algorithms are explained in detail in the supplementary methods section along with the parameter values used. The parameters for the tracking algorithms are hard-coded in TIAM. But we have provided information in the user guide as to where in the code the parameter values can be changed if desired. Information specific to the image series can be specified through the graphic user interface in order to calculate the motility characteristics of cells (see user guide).

2.3. Feature extraction and data integration

TIAM is designed to make use of the multi-channel image series in order to extract additional information on tracked cells to facilitate integrative analysis and provide insights into T cell motility. The feature extraction algorithms implemented in TIAM aim to retrieve physical features such as the area of attachment to some underlying substrate (from the reflection channel), polarity (from the transmitted light channel), and fluorescence intensity (from up to two fluorescence channels), and store/report them along with motility characteristics such as the cell's speed, turn angle, arrest coefficient, and confinement index (see Supplementary methods for description). The user interface provides options to specify the channels and the features to be extracted (see TIAM user guide). Due to the consistent perspective for all image channels, tracking results from the transmitted light image channel can be directly associated with secondary channels. The centroid of cells inferred at the detection step is used to link local pixel information from these secondary channels to the tracks ( Fig.ਁ ).

Discerning the boundary contour of a given cell is a common routine that is applied to any of the image channels, which can be defined as the Region of Interest (ROI) to calculate the desired features from that image channel. Given a centroid position, a square box of a pre-determined size around the centroid is used to isolate and select the local image. This local image ideally contains only the cell of interest. For the reflection and fluorescence channels, the local image is segmented via Otsu's method (Otsu, 1979) to give the cell boundary in that channel. In order to discard pixels associated with portions of touching neighboring cells, the Watershed algorithm (Meyer, 1994) is used on the distance transform of the initial segmented image. For the transmitted light channel, Canny edge detection (Canny, 1986) is used first to discern cell boundaries in the local image. In order to discard pixels associated with portions of touching neighboring cells, the Watershed algorithm is used on the CHT of the edge image. The largest region defined by the Watershed algorithm whose centroid is within a given distance from the center of the box is considered as the cell of interest.

The local segmentation approach was primarily implemented to handle reflection image series that tend to have spatiotemporally varying foreground and background pixel intensity values, which precludes the use of global thresholding. In addition, we found during the process of implementation that the Watershed algorithm was more reliable on the local images than the global images.

2.4. Additional features in TIAM

TIAM allows for batch processing of experimental datasets and can automatically distinguish the cell types based on differential fluorescent vital dye-labels (see Supplementary methods and user guide). TIAM also provides the option of having the selected image channel with the outlines of cells overlaid in a tiff image series. This can provide a visual assessment of the quality of segmentation of individual cells in that channel. A stand-alone MATLAB based user interface is provided to visualize individual or pairs of tracks in the video-mode (see user guide). This allows for manual inspection of tracking results from TIAM. This user interface is also intended to help in manually recording the track and frame numbers of desired corrections in track assignments. TIAM also provides a stand-alone track-editing feature that uses the manually compiled lists of desired corrections in track assignments (see user guide). The track-editing algorithm is a two-step process, where tracks are first split at specified frames (Fig. S4). Then the specified tracks and/or sub-tracks, either resulting from breakages in the first step or the ones that were missed by the algorithm, are joined together. Icy, an open source image analysis platform, also provides a plug-in for viewing and editing tracks (de Chaumont et al., 2012).

2.5. Evaluation of performance of detection and tracking

Performance evaluation, also referred as performance analysis, in image analysis compares the results obtained from an automated procedure against the manually established ‘ground truth’. Herein, a ground truth track represents the ‘true’ positions of a cell as a sequence of bounding boxes. We used the Video Performance Evaluation Resource (ViPER) software (Doermann and Mihalcik, 2000) to manually draw bounding boxes around cells in each video frame and index the sequences of bounding boxes corresponding to each individual cell to designate tracks.

Performance evaluation metrics were employed to quantitatively and comprehensively assess the detection and tracking performance of TIAM and the third-party tools. We used the Sequence Frame Detection Accuracy (SFDA) and Average Tracking Accuracy (ATA) metrics (Kasturi et al., 2009) as these can be computed in a fully automated fashion and thus allow for reproducible quantification of the success of detection and tracking of objects. Further, they do not suffer from the risk of human error or bias. These metrics have been adopted as standardized metrics by the Video Analysis and Content Extraction (VACE) program ( and the Classification of Events, Activities, and Relationships (CLEAR) consortium ( which are two large-scale and community-wide efforts concerned with video tracking and interaction analysis. The metrics are based on Jaccard Similarity (Fig. S5 for intuitive illustration and and Supplementary methods for mathematical description). In order to compute SFDA and ATA, a one-to-one correspondence between ground truth and result must be established. To establish this mapping we employed the Hungarian algorithm (Munkres, 1957) with metrics based on Jaccard Similarity used to construct the similarity matrix (see Supplementary methods for details).

We have consolidated the software routines to carry out performance analysis in a separate MATLAB-based suite that we call PACT (Performance Analysis of Cell Tracking). The PACT code, its user guide and relevant ground truth datasets are available at The user guide also includes specific instructions on using ViPER for ground truth annotation.

2.6. Evaluation of performance of feature extraction

Performance of feature extraction was also evaluated against ground truth. Outlines drawn manually or by semi-automated procedures in ImageJ (Schneider et al., 2012) were listed as ROIs and used as ground truth (see Supplementary methods for details). A one-to-one correspondence between individual cells in ground truth and TIAM result was obtained using the Hungarian algorithm (Munkres, 1957). The similarity matrix for the Hungarian algorithm was constructed for each frame using the distance between centroids of every possible pairing of cells in TIAM result with those in the ground truth. Once the one-to-one correspondence is achieved, the quantified features obtained from ground truth were compared against those from TIAM.


ICAnet overview

ICAnet is a module-based single-cell RNA-seq analysis tool, designed for integration, clustering and network analysis. This tool integrates gene expression information and high-quality PPI network in a novel way to precisely recover the landscape of single-cell expression atlas. ICAnet consists of three main steps: (i) Gene expression matrix preprocessing and decomposition (ii) Cross-batch expression programs clustering and (iii) Walk-trap-based activated ‘sub-network’ (module) identification. The details of these major steps are described as below.

Gene expression matrix preprocessing and decomposition

Gene expression normalization

ICAnet requires single-cell gene expression matrices as input, which are then normalized through a standard pre-processing step (log-normalization for all gene expression matrices using the size factor 10 000 per cell, log2CP10K). Users can also specify other types of gene expression quantification (e.g. TPM or RSEM) and normalization methods (e.g. SCTransfrom) before running the subsequent core steps of ICAnet.

Denoising gene expression matrix

In each dataset used for integration or clustering, ICAnet aimed to identify biological signals from gene expression matrices and to identify shared expression patterns. For different batches of datasets with different levels of data sparsity, the variability of the data sparsity will adversely affect comparisons of expression programs across different datasets, because part of data variation (signal) is driven by the data sparsity, not the actual biological signal ( 30). To diminish interference from data sparsity, ICAnet implemented two alternative strategies: (i) Computing top K variable genes for each batch according to the coefficient of variation for each gene, taking the intersection set of all sets of variable genes as the filtered gene set, and using their corresponding expression profile to perform ICAnet (ii) Using a recently developed Python module (named randomly) based on random matrix theory to denoise the dataset ( 30), which works very efficiently in eliminating single-cell sparsity-driven signals ( 30). We used it to denoise the dataset at first to prevent the influence of data sparsity on the matrix decomposition of ICAnet. In this study, we only applied the denoising preprocessing step to the pancreatic islet scRNA-seq datasets to improve the batch effect correction performance of ICAnet, because these datasets were generated from different library types and each dataset had different degrees of data sparsity.

Biological signal extraction via independent component analysis

To identify the biological signals (expression programs) in the dataset, we used ICA to decompose gene expression matrices into gene expression programs. The number of expression programs is a very important parameter in ICAnet, thus we proposed an unsupervised method based on random matrix theory ( 31, 32) to determine this parameter (see Supplementary Notes, Section 1 in the Supplementary Materials). Each dataset was centered before performing ICA for matrix decomposition. Two different implementations of ICA can be utilized by ICAnet. The first implementation is the joint approximate diagnalization of eigenmatrices (JADE) ( 33). The major advantage of JADE over other implementation solutions is that it is based on matrix computations involving matrix diagonalization, resulting in non-stochastic components. Other algorithms (e.g. FastICA) rely on an optimization procedure (e.g. starting points and optimization paths) ( 34), therefore, may yield variable results. The second implementation is based on the R package MineICA ( 25), which uses the same strategy as Icasso, to alleviate the stochastic problem when running FastICA ( 35) through iterative component clustering. In this study, we used JADE-based ICA to decompose gene expression matrices into independent components (source matrix), and the gene weights (importance) of each component have unit variance and zero means.

Cross-batch expression programs grouping

Grouping expression programs across batches to find shared biological signals

One key feature of ICAnet is the grouping of independent components (or expression programs) across different datasets/batches. First, ICA was performed independently on each dataset/batch. Then, the independent components computed from two (or more) single-cell datasets were compared by computing Pearson's correlation coefficient between corresponding gene weights of selected genes (projection value > 2.5 standard deviations in the identified component). After grouping of the components from different datasets/batches, Partitioning Around Medoids (PAM) algorithm ( 36) with the average silhouette width was used to estimate the optimal number of expression patterns. Finally, the medoids were chosen as the ‘basal programs’ shared across batches for further network weighting.

Activated ‘sub-network’ (module) identification

Construction of weighted PPI networks with basal programs shared across batches

In the following step, we combined PPI networks and expression programs to integrate their information. The PPI networks were obtained from the STRING database, a common and widely used PPI database ( 37). In this analysis, we used a threshold of a combined interaction score >600 to filter interactions, which is also a commonly used criterion for obtaining credible PPI networks ( 12, 38).

Those genes that significantly contribute to each expression program have been defined previously as the ‘activated genes’, which are identified using a weight threshold of three or four standard deviations from the mean. Here, we constructed weighted PPI network to produce activated sub-networks (or modules), wherein the edge-weight density is significantly greater than the rest of the network. We used the same weight scheme that used previously in computational epigenome model research ( 39). Specifically, for each component, the absolute weight value of each gene was determined and defined as ICA statistic |$(IC)$|⁠ . Assuming genes g and h are connected in the PPI, we assigned the edge weight as the average of the individual node (or gene) statistics, i.e. |$<>> = frac<1><2> ( <>+ IC> )$|⁠ . To avoid prohibitive computational expenditures, we only assigned the edge weights to the edges with endpoint ICA statistics that passed the weight threshold and zero was assigned to other edges. The weight threshold can be manually adjusted, and in this analysis, we set it as 2.5 standard deviations from the mean.

Random walk trapping to identify sub-networks in weighted PPIs

To rapidly and robustly identify dense connected and activated sub-networks, we used the random walk approach ( 40) to decipher all the possible sub-networks (modules). We performed random walks of different lengths using our ICA statistics-weighted PPI networks and detected modules by applying walk-trap algorithm on each random walk-based distance matrix. All the detected modules greater than three were saved and pooled together as module sets. We then applied the AUCell algorithm to the raw single-cell datasets to construct activated module–cell matrix that calculates the enrichment of each module in each cell as an area under the recovery curve (AUC) across the expression value-based rankings of all or some of the genes. The cell–module activity is summarized in a matrix (termed as module activity matrix) wherein columns represent single cells and rows represent the predicted modules.

Evaluation of clustering performance

Adjusted Rand Index (ARI)

When cell labels and batch information are available, the ARI can be used to calculate the similarity between the ICAnet clustering result and the known cell or batch labels (see Supplementary Notes, Section 2 in the Supplementary Materials).

Inverse Simpson's Index (LISI)

We used a score metric, named as LISI, to measure local diversity based on local neighborhood distribution (See Supplementary Notes, Section 2 in the Supplementary Materials). This index represents the expected number of cells that need to be sampled before neighboring cells are drawn from the same batch. The greater the score, the stronger the local batch_ID (iLISI) or cell_type (cLISI) heterogeneity is.

Clustering methods for cell states identification

For the cell states identification benchmark task, several methods were systemically compared. Before running clustering methods, we used count per million to derive a normalized count matrix. For t-Distributed Stochastic Neighbor Embedding (t-SNE)+k-means, pcaReduce and SC3, we used a log-transformed dataset and adjusted the number of clusters to optimize the clustering performance, which was evaluated by the ARIcell type. For SINCERA, we used z-score normalized data for the clustering analysis, and we also adjusted the number of expected clusters to optimize the ARIcell type values. For Seurat, we used the Seurat packages and processed related datasets in accordance with the tutorial ( We then performed cell clustering multiple times using Louvain clustering with multi-level refinement algorithms on a shared-nearest-neighbor-based cell graph, during which we adjusted the parameter resolution for the maximal ARIcell type. Three module-based clustering methods, SCENIC, SCORE and ICAnet, were compared in this study. All these methods quantified module activity based on AUCell. We ran each method and used the same aucMaxRank parameters to derive a module-based activity matrix.

For each clustering method, we used two variable gene selection criteria: the Top 5000 genes with the largest coefficient of variation, and the whole gene set. We then performed the above variable gene selection steps separately to select the criterion that produced the best clustering performance. For each test dataset, we re-analyzed the identifying novel rare cell types using Louvain clustering with a multilevel refinement algorithm ( 7) on a shared-nearest-neighbor-based cell graph derived from the module activity matrix to infer cell expression state.

Clustering methods for multi-batch datasets integration

In benchmarking different multi-batch integration methods, we used Louvain clustering with multilevel refinement algorithms on a shared-nearest-neighbor-based cell graph for each method, and adjusted the resolution parameter to obtain the optimal ARIcell type value. We then calculated corresponding LISI, iLISI and ARIbatch values. Additionally, for methods that correct batch effects on the Uniform Manifold Approximation (UMAP) space but not on the gene expression or PCA space in our study [e.g. BBKNN(41)], we applied Hierarchical DBSCAN + UMAP to cluster cells, and adjusted the parameters minPts to optimize the cell-clustering performance for comparisons.

Identification of cell type-specific activated modules

To identify activated modules for each cell type, we first identified cell type-associated modules using a receiver operating characteristic (ROC) curve analysis ( 7). For each gene, we evaluated a classifier that was built on that module alone, to distinguish a specific group of cells from other cells. An AUC value close to 1 indicates that this module is more specifically expressed in a specific cell group. We implemented the above analysis using the FindMarker function provided by Seurat ( 7), with AUC > 0.75 as a threshold to call cell type-associated modules. Then, among the cell type-associated modules, continuous module activity was converted into binary values using AUCell ( 11) and the Spearman's correlation coefficient between each cell type and the binarized module were calculated. The modules with Spearman's coefficient < 0.3 were filtered out. Finally, the resulting modules with statistical significances greater than the threshold (P-value < 0.05, see Supplementary Notes, Section 5 in the Supplementary Materials) were selected and defined as cell type-specific activated modules.

Stability and robustness evaluation of three module-based clustering algorithms

To test the stability of three module-based clustering algorithms [ICAnet, SCENIC ( 11) and SCORE ( 12)], we performed two different tests: (i) down-sampling the datasets with varied cell numbers (2000, 1000, 500 and 100) and (ii) simulation of low-sequencing depth by reducing the expression level to one-fifth of the original. We used the same down-sampling and gene expression simulation procedures for all the three tested methods, and the tSNE+DBSCAN clustering algorithm was performed to evaluate the newly predicted clusters. Finally, we calculated the ARI between the labels of identified clusters and previously annotated cell-type labels. In the clustering step, we ran DBSCAN multiple times, during which we altered the parameter epsilon in the range of 1.0–4.0 and minPts in the range of 1–50 to determine a maximal ARI.

Module recovery analysis

For the ICAnet-weighted PPI network, a K number of different weighted PPI networks were determined. An over estimation of the number of weighted PPI networks results in some false positives during module recovery therefore, we only computed the first independent component and created corresponding weighted PPI networks for the downstream analysis.

Label-association analysis using graph signal processing

To identify which is the novel cell type (or state) among our cell-type labeling results, the intrinsic ‘label association’ between our cell-type annotations and those defined by the original author need to be determined. Inspired by a recently proposed signal-enhancing model ( 44), we used graph signal smoothing to transform the binary ‘cell-type label signal’ into a continuous ‘cell-type label signal’ to enhance label association.

Therefore, the y is the reconstructed continuous signal vector for cell type i. We applied graph smoothing to each cell type to derive their continuous signal vector, and calculated the Pearson's correlation matrix between our annotated cell types and those in the raw cell-type annotations. β was assigned a value of 0.8 in this step. Furthermore, we used cor’ = (1+cor)/2 to transform the correlation matrix, and used 0.6 (Pearson's correlation coefficient > 0.2) and the FDR (false discovery rate) < 0.05 as thresholds to identify significant associations.

Gene set enrichment analysis

We used the software GSEA (version 4.1.0), a Java desktop application to assess potential enrichment of specific gene sets in a ranked list of differentially expressed genes for each cell type. The curated gene sets are consistent of cancer stemness/risk associated gene-sets ( 45) and AML risk-gene BAALC expression associated gene-set ( 46).

Survival analysis of acute myeloid leukemia (AML) patient based on module activity

To measure the activity levels of modules inferred from the scRNA-seq datasets in bulk RNA expression datasets, we first used gene set variation analysis (GSVA) ( 47) to calculate the module activity in each bulk sample. After converting the gene expression matrix into a module activity matrix, we selected the best subset of modules to predict survival in the training cohort. We used a linear regression model named Least Absolute Shrinkage and Selection Operator (LASSO) implemented by the glmnet R package ( 48). By enabling a 10-fold cross-validation to fit a Cox regression model, we were able to identify an optimized set of modules to predict survival. Owing to the randomness of the LASSO model, we applied a bootstrapping strategy to score each module. This procedure generated 100 resampled datasets from the complete sample sets, with a sample size equal to 80% of the whole samples. LASSO was performed with 10-fold cross validation to optimize the parameters for module selection in each resampled dataset. Finally, we scored each module based on how frequent this module was selected by the regression model during bootstrapping. On the basis of the resulting scores, we selected the top-K modules and performed PAM clustering on the samples guided by the selected feature modules to predict patient survival. We used the Top30 modules as AML patient-associated modules, because they yielded the most significant patient survival difference in the training dataset.

Materials and methods

Fermentation analysis

Two cultures of C. acetobutylicum ATCC 824 were grown in pH controlled (pH >5) bioreactors (Bioflow II and 110, New Brunswick Scientific, Edison, NJ, USA) [7]. Cell density, substrate and product concentrations were analyzed as described [56].

RNA isolation and cDNA labeling

Samples were collected by centrifuging 3-10 ml of culture at 5,000×g for 10 minutes, 4°C and storing the cell pellets at -85°C. Prior to RNA isolation, cells were washed in 1 ml SET buffer (25% sucrose, 50 mM EDTA [pH 8.0], and 50 mM Tris-HCl [pH 8.0]) and centrifuged at 5,000×g for 10 minutes, 4°C. Pellets were processed similarly to [7] but with the noted modifications. Cells were lysed by resuspending in 220 μl SET buffer with 20 mg/ml lysozyme (Sigma, St. Louis, MO, USA) and 4.55 U/ml proteinase K (Roche, Indianapolis, IN, USA) and incubated at room temperature for 6 minutes. Following incubation, 40 mg of acid-washed glass beads (≤106 μm Sigma) were added to the solution, and the mixture was continuously vortexed for 4 minutes at room temperature. Immediately afterwards, 1 ml of ice cold TRIzol (Invitrogen, Carlsbad, CA, USA) was added 500 μl of sample was diluted with an equal volume of ice cold TRIzol and purified. Following dilution, 200 μl of ice cold chloroform was added to each sample, mixed vigorously for 15 s, and incubated at room temperature for 3 minutes. Samples were then centrifuged at 12,000 rpm in a tabletop microcentrifuge for 15 minutes at 4°C. The upper phase was saved and diluted by adding 500 μl of 70% ethanol. Samples were then applied to the RNeasy Mini Kit (Qiagen, Valencia, CA, USA), following the manufacturer's instructions. To minimize genomic DNA contamination, samples were incubated with the RW1 buffer at room temperature for 4 minutes. The method disrupted all cell types equally, as evidenced by microscopy (data not shown). cDNA was generated and labeled as described [7]. The reference RNA pool contained 25 μg of RNA from samples taken from the same culture at 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 44, 48, 54, 58, and 66 h.

Microarray analysis

Agilent technology 22k arrays, (GEO accession number GPL4412) as described in [63], were hybridized, washed, and scanned per Agilent's recommendations. Spot quantification employed Agilent's eXtended Dynamic Range technique with gains of 100% and 10% (Agilent's Feature Extraction software (v. 9.1)). Normalization and slide averaging was carried out as described [7, 63]. A minimum intensity of 50 intensity units was used as described [63]. Microarray data have been deposited in the Gene Expression Omnibus database under accession number GSE6094. To gain a qualitative measure of the abundance of an mRNA transcript, the averaged normalized log mean intensity values were ranked on a scale of 1 (lowest intensity value) to 100 (highest intensity value). Genes were clustered using TIGR's MEV program [64].

Quantitative RT-PCR

Q-RT-PCR was performed as described [48]. Specific primer sequences are included in Additional data file 9 CAC3571 was used as the housekeeping gene.


For light microscopy, samples were stored at -85°C after 15% glycerol was added to the sampled culture. Samples were then pelleted, washed twice with 1% w/v NaCl and fixed using 50 μl of 0.05% HCl/0.5% NaCl solution to a final count of 10 6 cells/μl. Slides were imaged using a Leica widefield microscope with either phase contrast or Syto-9 and PI dyes (Invitrogen LIVE/DEAD BacLight Kit) to distinguish cell morphology.

For electron microscopy, samples were fixed by addition of 16% paraformaldehyde and 8% glutaraldehyde to the culture medium for a final concentration of 2% paraformaldehyde and 2% glutaraldehyde. For cultures grown on plates, colonies were scraped from the agar and suspended in 2% paraformaldehyde and 2% glutaraldehyde in 0.1 M sodium cacodylate buffer (pH 7.4). Cultures were fixed for 1 h at room temperature, pelleted and resuspended in buffer.

For transmission electron microscopy, bacteria were pelleted, embedded in 4% agar and cut into 1 mm × 1 mm cubes. The samples were washed three times for 15 minutes in 0.1 M sodium cacodylate buffer (pH 7.4), fixed in 1% osmium tetroxide in buffer for 2 h, and then washed extensively with buffer and double de-ionized water. Following dehydration in an ascending series of ethanol (25, 50, 75, 95, 100, 100% 15 minutes each), the samples were infiltrated with Embed-812 resin in 100% ethanol (1:3, 1:2, 1:1, 2:1, 3:1 1 h each) and then several changes in 100% resin. After an overnight infiltration in 100% resin, the samples were embedded in BEEM capsules and polymerized at 65°C for 48 h. Blocks were sectioned on a Reichert-Jung UltracutE ultramicrotome and ultrathin sections were collected onto formvar-carbon coated copper grids. Sections were stained with methanolic uranyl acetate and Reynolds' lead citrate [65] and viewed on a Zeiss CEM 902 transmission electron microscope at 80 kV. Images were recorded with an Olympus Soft Imaging System GmbH Megaview II digital camera. Brightness levels were adjusted in the images so that the background between images appeared similar.

For scanning electron microscopy, fixed samples were incubated on poly-L-lysine coated silica wafers for 1 h and then rinsed three times for 15 minutes in 0.1 M sodium cacodylate buffer (pH 7.4). The samples were fixed with 1% osmium tetroxide in buffer for 2 h, washed in buffer and double de-ionized water, and then dehydrated in ethanol (25, 50, 75, 95, 100, 100% 15 minutes each). The wafers were critical point dried in an Autosamdri 815B critical point drier and mounted onto aluminum stubs with silver paint. The samples were coated with Au/Pd with a Denton Bench Top Turbo III sputter-coater and viewed with a Hitachi 4700 FESEM at 3.0 kV.

Phylogenetic tree generation

Based on the genome annotations available at NCBI, we considered any sigma factor that was annotated as σ 70 or unannotated. A second filter was applied by requiring that all the sequences should contain a Region 2, the most conserved region of the σ 70 protein. All members of this class of sigma factor contain Region 2, and it was modeled with the HMM pfam04542. This criterion removed CAC0550, CAC1766 and CAP0157, but they were added to the list again despite their lack of a Region 2. The alignment was made using ClustalW 1.83 using the default settings and visualized as a radial tree as created by Phylodraw v. 0.8 from Pusan National University.

Generation and characterization of antisense strains

Oligonucleotides were designed to produce asRNA complementary to the upstream 20 bp and first 30-40 bp of the targeted genes' transcripts (Additional data file 7). The constructs were cloned into pSOS95del under the control of a thiolase (thl) promoter and confirmed by restriction digest. Plasmids were then methylated and transformed into C. acetobutylicum ATCC 824, as previously described [33, 55, 56]. Strains were grown in 10 ml cultures and characterized using microscopy and HPLC to analyze final product concentrations [56].


By doing this course well, students will develop basic knowledge and skills in cell and molecular biology and become aware of the complexity and harmony of the cell. As students proceed through the modules, they will be able to apply this knowledge, skill, and awareness to topics like the following:

  • Basic properties of cells
  • Prokaryotic and eukaryotic cells
  • Viruses
  • Biological molecules: carbohydrates, lipids, proteins, and nucleic acids
  • Techniques used in cell and molecular biology
  • Enzymes
  • Metabolism
  • Mitochondrion structure and function
  • Chloroplast structure and function
  • Plasma membrane composition, structure, and function
  • The movement of substances across cell membranes
  • The endomembrane system
  • The extracellular matrix
  • The structure and function of the nucleus
  • Genes and chromosomes
  • DNA replication
  • Transcription
  • Translation
  • Cytoskeleton and cell motility
  • Cellular reproduction
  • Cell signaling
  • Cancer


Semen cryopreservation is a well-established procedure used in veterinary assisted reproduction technology applications. We investigated damaging effects of cryopreservation on the structural and ultrastructural characteristics of bull sperm induced at different temperatures and steps during standard cryopreservation procedure using transmission (TEM) and scanning electron microscopy. We also examined the effect of cryopreservation on sperm DNA and chromatin integrity. Five healthy, fertile Friesian bulls were used, and the ejaculates were obtained using an artificial vagina method. The semen samples were pooled and diluted in a tris-yolk fructose (TYF) for a final concentration of 80 × 10 6 spermatozoa/ml. The semen samples were packed in straws (0.25 ml), and stored in liquid nitrogen (−196°C). Samples were evaluated before dilution, just after dilution (at 37°C), at 2 h and 4 h during equilibration, and after thawing (37°C for 30 s in water bath). In association with step-wise decline in motility and viability, our results showed that the plasma membrane surrounding the sperm head was the most vulnerable structure to cryo-damage with various degrees of swelling, undulation, or loss affecting about 50% of the total sperm population after equilibration and freezing. Typical acrosome reaction was limited to 10% of the spermatozoa after freezing. We also observed increased number of mitochondria with distorted cristae (15%). Chromatin damage was significantly increased by cryopreservation as evident by TEM (9%). This was mainly due to DNA breaks as confirmed by Sperm Chromatin Structure Assay (SCSA) (8.4%) whereas the chromatin structure was less affected as evaluated microscopically by toluidine blue staining. We concluded that, using standard cryopreservation protocol, the most pronounced damage induced by cryopreservation is observed in the plasma membrane. Further improvement of cryopreservation protocols should thus be targeted at reducing plasma membrane damage. Acrosomal, mitochondrial and chromatin damage are also evident but appear to be within acceptable limits as discussed.


Michael Ailion | Neuromodulator release and signaling at the cellular and organismal levels.

Chip Asbury | Molecular basis of mitosis.

Jihong Bai | How synapses are assembled into functional circuits using a combination of genetic, biochemical, imaging, and electrophysiological methods..

Wyeth Bair | Computer modeling and electrophysiology of the visual system.

Sandra Bajjalieh | Cell biology of neurons with emphases on membrane trafficking and lipid signaling pathways.

Melissa Barker-Haliski | Preclinical models of epilepsy and mechanisms of epileptogenesis in the elderly.

Andres Barria | Molecular mechanisms controlling synaptic function and plasticity. Role of NMDA receptors.

Michael Beecher | Auditory communication in birds

Olivia Bermingham-McDonogh | Mechanisms of development and regeneration of the mammalian auditory system.

Marc Binder | Properties of voltage-gated membrane channels.

Martha Bosma | Development of central nervous system neurons using physiological and molecular techniques.

Mark Bothwell | Growth factor mechanisms in neural development and degenerative disease.

Geoffrey Boynton | Functional organization of human visual perception.

Eliot Brenowitz | Neural basis of biologically relevant behavior in animals, and the cellular and molecular mechanisms of plasticity in adult brains.

Michael Bruchas | Dr. Bruchas’ laboratory focuses on understanding how brain circuits are wired, how they communicate with one another, and dissecting the neural basis of stress, emotion and reward.

Bingni Brunton | Data-driven low-dimensional dynamic models of neuronal networks.

Linda Buck | Odors, tastes, pheromones, stimulating specific behaviors or physiological effects in conspecifics.

Steven Buck | Focuses on human color vision and linking perceptual visual experience and the underlying neural/genetic substrate.

Elizabeth Buffalo | Our research is aimed at understanding the neural mechanisms that support learning and memory.

Clemens Cabernard | The Cabernard lab is studying asymmetric cell division (ACD), a process that generates cellular diversity.

Erik Carlson | The study of cerebro-cerebellar circuits in mice relevant to neuropsychiatric and neurodegenerative disease.

Steven Carlson | Synapse formation – focusing on the formation of the active zone, the site of neurotransmitter release in the nerve terminal.

William Catterall | Molecular basis of electrical excitability molecular and cellular biology of ion channels function of calcium channels in neurotransmission.

Charles Chavkin | Using mouse genetic and optical stimulation of CNS pathways, we study how stress exposure affects depression, drug addiction risk, and cognition by affecting neural circuits and molecular signaling.

Daniel Chiu | The development of new tools, based on nanmaterials, optics, and microfluidics, for interfacing and interrogating neuronal systems and synaptic function at the nanometer scale.

Howard Chizeck | Biorobotics, telerobotics and neural engineering.

Eric Chudler | Cortical and basal ganglia mechanisms of nociception and pain, the neuroactive properties of medicinal plants and herbs, and translating basic neuroscientific research into language and activities for the general public.

John Clark | Characterizing the functional mechanism for the protective actions of the stress protein, human alphaB crystallin, a lens protein that is upregulated in aging diseases and protects against protein unfolding/misfolding and aggregation in Alzheimer’s, Huntington’s, Parkinson’s disease and cataracts.

David G. Cook | Molecular mechanisms of neurodegenerative disorders.

Mark Cooper | Gastrulation and neurulation in zebrafish embryos cell motility.

Ellen Covey | Structure and function of the central auditory system and the neural basis of echolocation neural mechanisms for processing temporal patterns of sound.

Dennis Dacey | Structure and function of the primate retina.

Valerie Daggett | Molecular modeling of proteins implicated in disease. Design and testing of diagnostic and therapeutic agents for neurodegenerative diseases.

Raimondo D’Ambrosio | Pathophysiology of glial cells and basic mechanisms of epilepsy. Specific current interest include glial extracellular ion homeostasis in traumatic brain injury, stroke and posttraumatic epilepsy membrane potassium channels edema.

Thomas Daniel | Sensorimotor control of animal movement.

Marie Y. Davis | The goal of my research is to understand mechanisms causing neurodegeneration in human movement disorders.

Horacio de la Iglesia | Neural basis of circadian behavior.

Nikolai C. H. Dembrow | How dendritic integration and the neuromodulation of intrinsic neuronal properties in distinct neuron types shapes neocortical function.

Peter Detwiler | Signal transduction in retinal photoreceptors.

Ajay Dhaka | Biology of somatosensation via molecular, cellular, developmental and behavioral investigation.

Jaime Diaz | Disruptions of the growth program affecting adult brain function and how other metabolic systems function.

Adrienne Fairhall | Computational approaches in neuroscience: adaptive and multimodal sensory processing, biophysics of computation by single neurons and small circuits, algorithms of computation in diverse systems .

Angela Fang | neurobiological correlates of maladaptive social cognition in anxiety and obsessive-compulsive related disorders to inform personalized treatment prediction role of oxytocin in the pathophysiology of these disorders.

Susan Ferguson | Using novel viral vector methods to unravel the role of cortico-basal ganglia circuitry in the development of behaviors that contribute to drug addiction, as well as in the processes that regulate decision-making, motivation and impulsivity.

Eberhard Fetz | Properties of cortical and spinal neurons controlling limb movement in primates dynamic neural network modeling implanted recurrent brain-computer interfaces.

Ione Fine | Effects of long-term visual deprivation, perceptual learning and plasticity, psychophysics, fMRI and computational vision.

Albert Folch | Neurobiology on a chip (Neuro-MEMS): Microengineered systems to study synaptogenesis, axon guidance, ion channel activity, and olfaction, among other neuroscience topics

Stanley Froehner | Molecular basis of synapse formation and function.

Jose M. Garcia | My research focuses on neuroendocrinologic aspects of traumatic brain injury and on hormonal pathways in wasting conditions.

David Gire | How neural circuits process natural spatiotemporal olfactory sensory cues to guide flexible, ethologically relevant behaviors.

Sam A. Golden | Neurobiology and circuitry of affective social behaviors and neuropychiatric disease.

Sharona Gordon | Ion channel biophysics & trafficking and regulation of neuronal plasticity in sensory transduction.

Thomas Grabowski | Functional magnetic resonance imaging studies of the neural systems basis of language and cognition in health and disease.

Brock Grill | Proteomic and genetic interrogation of neuronal signaling.

Chris Hague | Functional characterization of adrenergic receptors.

Julie Harris | Understanding the relationship between anatomical and functional neural circuitry between brain areas in normal and disease states.

Jeffrey Herron | Developing new research tools and systems to explore the applications of bi-directional neural interfaces to enable or improve the treatment of neurological diseases, disorders, and injuries.

Bertil Hille | Modulation of ion channels by G protein coupled receptors and membrane phosphoinositide lipids.

Greg Horwitz | Visual perception and viral vector-mediated gene transfer in primates.

Clifford Hume | Mammalian inner ear development, gene therapy of inner ear disorders, and imaging analysis of the inner ear.

James Hurley | Mechanisms of phototransduction light and dark adaptation.

Sandra Juul | Developing neuroprotective strategies for infants at high risk for neurodevelopmental impairment using in vivo and in vitro models of preterm and term brain injury.

Brian Kalmbach | Neurophysiology of primate neocortical cell types. Ion channels, neuromodulators and related genes.

Franck Kalume | Investigations of mechanisms and treatments of genetic epilepsies in animal models.

Jeansok Kim | Neurocognitive effects of stress basic mechanisms of fear.

Natalia Kleinhans | Multimodal imaging and neuropsychological assessment of neuropsychiatric disorders.

Andrew Ko |My research focuses on human electrophysiological and imaging correlates of behavior, disease, and interventions for epilepsy, movement disorders and pain.

Brian Kraemer |Molecular causes of neurodegeneration in Alzheimer’s disease, amyotrophic lateral sclerosis (ALS), and related disorders of the nervous system.

Patricia Kuhl | Speech perception throughout the lifespan with an emphasis on early development behavioral as well as ERP, fMRI, and MEG studies on language processing.

Adrian KC Lee | Auditory brain sciences and neuroengineering.

Ed Lein | Molecular, cellular and circuit organization of the developing and adult human neocortex.

Nicole Liachko | Seeks to understand the biology underlying neurodegenerative diseases of aging including Amyotrophic Lateral Sclerosis (ALS) and Alzheimer’s disease.

Michael Manookin | Neural circuits, cells, and synapses that mediate early visual processing.

Ludo Max | The role of sensorimotor integration in human motor learning and motor control, with an emphasis on auditory-motor learning in speech production and visuo-motor integration in limb movements.

G. Stanley McKnight | The role of intracellular signaling systems in the neuronal circuits that affect feeding and energy metabolism.

Kathleen Millen | The Millen laboratory uses molecular genetic approaches to explore the pathogenesis of congenital birth defects of the human and mouse brain and to study genes essential for normal neurodevelopment.

Dana Miller | We use C. elegans to understand how changes in environmental conditions are integrated into organism physiology, and how these responses modulate cellular processes involved in neurodegeneration.

Sheri J.Y. Mizumori | Neurobiology of learning and memory.

Cecilia Moens | Developmental genetics of brain patterning in the zebrafish.

William Moody | The role of spontaneous activity in cortical development, with some emphasis on the basic mechanisms underlying pediatric epilepsy.

Randall Moon | Functions and mechanisms of action of the wnt signaling pathways in embryonic development, regeneration, and diseases, and development of therapies to treat these diseases based on high throughput small molecule screens and genome-wide RNAi screens.

Claudia Moreno | Molecular mechanisms of aging.

Chet Moritz | We are developing neuroprosthetic technology for the treatment of paralysis and other movement disorders.

Gabe J. Murphy | We determine how particular synapses, cells, and circuits organize and extract the information that enables visually-guided behavior.

Scott O. Murray | Understand the brain mechanisms and cognitive process by combining behavioral and functional (fMRI) measurements of neural activity.

Neil M. Nathanson | Regulation of expression and function of muscarinic and neurokine receptors.

Jay Neitz | Biology of vision and vision disorders.

Maureen Neitz | Biology of vision and vision disorders.

John Neumaier | Molecular, cellular, and circuitry aspects of stress and addiction.

Jeffrey Ojemann | Electrocorticography studies of cognition and brain-computer interface.

Jaime Olavarria | Structure, function, and development of topographically organized circuits in the mammalian visual system.

Shawn Olsen | Cortical mechanisms of visual behavior and cognition.

Amy Orsborn | Engineering and understanding learning to develop neural interfaces

Lee Osterhout | Psychological and neural underpinnings of human language psychophysiological studies of human language and memory.

Leo Pallanck | Genetic analysis of neurotransmitter release mechanisms in Drosophila.

Richard Palmiter | Our laboratory uses mouse genetic models and viral gene transfer to dissect neural circuits involved in innate behaviors.

Jay Parrish | We are broadly interested in understanding the form and function of somatosensory neurons in Drosophila.

Anitha Pasupathy | Neural basis of visual shape representation and recognition in the primate brain.

David Perkel | Neural mechanisms of learning, focusing on vocal learning in songbirds anatomical and electrophysiological techniques for study of neuronal processing related to behavior.

Steve I. Perlmutter | Understanding and manipulating neural plasticity in mammalian motor systems to develop new therapies that improve recovery after spinal cord injury and brain damage.

Paul Phillips | The role of rapid dopamine neurotransmission in motivated behavior and decision making, and its dysfunction in mental health disorders including addiction.

Nicholas Poolos | Mechanisms of ion channel dysregulation in epilepsy, and impact on dendritic excitability.

Chantel Prat | My research investigates the biological basis of individual differences in language and cognitive abilities.

Daniel Promislow | We use quantitive genetics and systems biology to identify naturally occurring modifiers of neurodegenerative disease in the fruit fly.

David Raible | Zebrafish mechanosensory hair cell development, damage and regeneration: models for hearing loss.

Akhila Rajan | Fat-brain communication: how fat signals to the brain.

Jan-Marino (Nino) Ramirez | Understanding the neuronal basis of a variety of brain functions to find novel ways to treat and cure neurological disorders in children, including epilepsy, Rett syndrome, brain tumors, and sudden infant death syndrome.

Rajesh P.N. Rao | Computational neuroscience, machine vision and robotics, and brain-computer interfaces.

Wendy Raskind | The genetic etiologies of Mendelian neurodegenerative disorders, including ataxias and parapareses, and the complex neurobehavioral disorder dyslexia.

Jeff Rasmussen | Molecular and cellular regulation of zebrafish somatosensory neuron development and repair

Tom Reh | Determination of the mechanisms that control neuronal proliferation and differentiation during neurogenesis of the vertebrate CNS.

R. Clay Reid | Deciphering how information is encoded and processed in neural networks of the visual system, using behavior, anatomy and physiology.

Fred Rieke | Visual signal processing and computation phototransduction.

Jeff Riffell | Olfactory neurobiology and chemical communication processes.

Farrel R. Robinson | Cerebellar control of movements using monkey eye movements as a model.

Ariel Rokem | Conducts research on the biological basis of brain function using computational tools that we develop and maintain.

Edwin Rubel | evelopment and habilitation of the inner ear and CNS auditory pathways.

Jay Rubinstein | Signal processing, physiology, and perception with inner ear implants using both computational modeling and experimental techniques.

Hannele Ruohola-Baker | Regulation of stem cell self renewal and regeneration.

Ramkumar Sabesan | Functional imaging of the human retina.

Abigail G. Schindler | Computational medicine, iterative translation, and systems biology to understand traumatic stress and its comorbidities.

John Scott | Specificity of synaptic signaling events that are controlled by kinase anchoring proteins.

Eric Shea-Brown | Computational and theoretical neuroscience.

Andy Shih | Our lab uses imaging approaches to better understand the regulation of brain microvascular health, and the factors that lead to its dysfunction during disease.

Joseph Sisneros | Understanding how the vertebrate auditory system processes species-specific vocalizations and the adaptive mechanisms that are used to optimize the receiver’s sensitivity to social communication signals.

Stephen Smith | The behavior of protein interaction networks at the neuronal synapse.

William Spain | Transformation of synaptic inputs into patterns of action potential output information flow within the network of neurons.

Kat Steele | Dynamics and control of human movement.

Nicholas A. Steinmetz | Distributed neural circuits underlying visually-guided behavior in mice.

Nephi Stella | Activation of immune cells in the CNS.

Jennifer Stone | Study of cellular and molecular mechanisms underlying generation of sensory hair cells in the inner ear during development, under normal conditions, and after injury.

Daniel Storm | Molecular and cellular basis of long-term memory and memory persistence using an interdisciplinary approach.

Garret Stuber | Research in the Stuber lab uses an interdisciplinary approach to study the neural circuit basis of motivated behavior.

Jane Sullivan | Cellular and molecular mechanisms controlling synaptic transmission and plasticity.

Billie Swalla | he evolution of chordates, especially the central nervous system. We are studying the gene networks that specify the central nervous system, in invertebrate deuterostomes and chordate embryos and adults.

Gregory W. Terman | Neurophysiology and pharmacology of synaptic plasticity in pain transmission pathways of the central nervous system as a model for the pathogenesis of chronic pain.

James Thomas | Molecular evolution, especially the evolution and function of gene families implicated in environmental interactions and other rapidly changing selective pressures. Work is mostly on nematode and mammalian gene families, with some comparative analyses to other groups.

Jonathan T. Ting | Biophysical, anatomical, and molecular features of human neocortical cell types and leverage this information to develop novel molecular genetic tools for accessing and perturbing brain cell types across diverse mammalian species.

Eric Turner | The mechanisms of brain development and neural gene regulation, and brain pathways affecting mood and anxiety. Using transgenic mouse models.

John Tuthill | Neural mechanisms of somatosensory processing and adaptive motor control.

Russell Van Gelder | Natural and synthetic inner retinal and extraocular photoreception.

Oscar Vivas | My lab uses electrophysiology and imaging to understand the changes in the autonomic nervous system during aging.

Jack Waters | Cells and circuits of the neocortex and their modulation with behavioral state, studied primarily with optical techniques.

Kurt Weaver | My research focuses on the dynamic interplay between large-scale neural systems and cognitive function, how this interaction can better inform contemporary models of neurological and psychiatric disorders.

Sara J. Webb | The use of physiological and neuroimaging methods to study attention, perception and social learning in children and adults with autism and other neurodevelopmental disorders

Jonathan Weinstein | The neuroimmune response in stroke and ischemic preconditioning (IPC) with emphasis on the role of type 1 interferon signaling in microglia in IPC-mediated endogenous neuroprotection.

John Welsh | Our work focuses on the role of neuronal oscillation in cognitive and motor function.

Rachel Wong | Circuit assembly and reassembly in the developing nervous system.

Zhengui Xia | The effect of genes and environmental exposure on adult neurogenesis, cognitive impairment, and neurodegeneration.

Libin Xu | The roles of lipid oxidation and metabolism in neurological diseases.

Smita Yadav | Elucidating the role of kinase signaling in neuronal development and disease using chemical-genetics, proteomics and stem cell techniques.

Azadeh Yazdan-Shahmorad | Developing novel neural technologies for neurorehabilitation.

Jason Yeatman | Quantitative neuroimaging of brain development and learning to read

Jessica Young | Building robust human models of Alzheimer’s disease.

Cyrus P. Zabetian | The genetics of neurodegenerative diseases with an emphasis on Lewy body disorders.

William N. Zagotta | Mechanisms of ion channel function.

Jing Zhang | Proteomics investigation of molecular mechanisms of Parkinson’s disease, and biomarker discovery for neurodegenerative diseases.

Larry Zweifel | Understanding the mechanisms of phasic dopamine-dependent modulation of reward and punishment, and the role of dopamine in generalized fear and anxiety.


Mitochondrial quality is under surveillance by autophagy, the cell recycling process which degrades and removes damaged mitochondria. Inadequate autophagy results in deterioration in mitochondrial quality, bioenergetic dysfunction, and metabolic stress. Here we describe in an integrated work-flow to assess parameters of mitochondrial morphology, function, mtDNA and protein damage, metabolism and autophagy regulation to provide the framework for a practical assessment of mitochondrial quality. This protocol has been tested with cell cultures, is highly reproducible, and is adaptable to studies when cell numbers are limited, and thus will be of interest to researchers studying diverse physiological and pathological phenomena in which decreased mitochondrial quality is a contributory factor.

Watch the video: Cell motility simulation (December 2022).