Apps & Workflows

Apps & Workflows

Apps & Workflows

FASTGenomics already provides different workflows for the analysis of single cell transcriptomic data. Thereby, each workflow consists of a set of FASTGenomics calculation and visualization apps. Below you can find a description of the currently available workflows, calculation and visualization apps.

Apps

  • 3D Scatter Plot

    Interactive 3D scatterplot. Different coloring schemes can be chosen to visualize cell cluster assignment, batch information and gene expression.

  • 3D Scatter Plot for large data sets

    This app provides an interactive scatterplot for more than one million cells.

  • App Template (minimal)

    This app contains a very minimal base-image for FASTGenomics calculations with faster startup-times.

  • App Template for Python 3.6

    This app contains a python-based image for FASTGenomics calculations with everything you need to start developing a calculation app.

  • App Template for Scanpy

    This app contains a python-based image for FASTGenomics calculations with scanpy-support. It contains everything you need to start developing a calculation app that makes use of the scanpy functionality.

  • Bar Plot

    This app produces a bar plot. Use this app e.g. to show predicted labels in a classification.

  • Batch Effect Check

    The Batch Effect Check identifies batch effects or similar technical artifacts based on an input of gene expression data and batch labels. A score for the presence of batch effects is computed based on the ability of a random forest classification trained to predict individual batch categories and compared with a dummy classifier. This app can be combined with the Batch Effect Report to visualize its results.

  • Batch Effect Report

    This app visualizes a selection of different statistics on batch effects. Use this app in combination with the according calculation app.

  • Classification Model Trainer

    The Classification Model Trainer App assigns cells to previously established classes based on their gene expression signatures. Different classification algorithms such as logistic regression, support vector machines, random forest classification and gradient boosting can be used to obtain the best classifier on the training data.

  • Clustering (h-DBSCAN)

    This Clustering App defines discrete cell subtypes in a data set by assigning cells of similar gene expression profiles to a common cluster. This app is based on the h-DBSCAN algorithm. To increase performance and clustering stability, you may want to combine this algorithm with a suitable dimensionality reduction method (e.g. truncated SVD).

  • Clustering (Hierarchical)

    This Clustering App identifies substructures in a single cell data set by assigning cells of similar gene expression profiles to a common cluster in a hierarchical fashion. This app is based on the average linkage algorithm.

  • Clustering (Louvain)

    This Clustering App defines discrete cell subtypes in a data set by assigning cells of similar gene expression profiles to a common cluster. This app is based on the Louvain algorithm.

  • Clustering (Seurat Adaptation by LIMES)

    This custom Clustering App performs preprocessing, variable genes selection, dimensionality reduction and clustering using functionality implemented in the Seurat package as typically applied in the LIMES Lab of Bonn University.

  • Confusion Map

    This app evaluates the stability of a given clustering to assess its quality. This is achieved by training and evaluationg a logistic regression classifier trying to predict the cluster assignments from the gene expression profile of each cell.

  • Confusion Map

    This app creates a confusion map created based on data provided in a CSV file.

  • Count Data Normalization (log)

    This Normalization App scales the gene expression count data to stabilize variance by applying a log transformation. Use this app when distance metrics are used in downstream analyses, e.g. when performing dimensionality reduction.

  • Count Data Normalization (tf-idf)

    This app normalizes gene expression counts to amplify the contribution of highly discriminative genes by using the TF-IDF scheme. Use this app as a noise reduction method before running dimensionality reduction tools.

  • Data Import

    This app can be used to import data of different formats and convert the into the internal FASTGenomics format.

  • Data Quality Report

    This app creates a data quality report providing basic insights on the count distributions per cells and genes.

  • Data Transformation for Scatter-HD

    This app performs a data transformation required for the Scatterplot HD visualization app.

  • Differential Gene Expression (Generalized Linear Models)

    This Differntial Gene Expression app finds marker genes with significantly different expression between different categories of cells, e.g. cell subtypes in a cell clustering or different cell branches in an infered cell differentiation. It fits a generalized linear model to the cell categories and performes a hypothesis test for significantly differing distributions.

  • Differential Gene Expression (Non-Parametric)

    This Differntial Gene Expression app finds marker genes with significantly different expression between different categories of cells, e.g. cell subtypes in a cell clustering or different cell branches in an infered cell differentiation. The app is based on a two-step procedure using a Kruskal-Wallis test for global group differences and Mann-Whitney-U tests between clusters. Use this in case you can make no prior assumptions on the distributions of gene expression levels.

  • Differential Gene Expression (Scanpy Diff Rank)

    This Differntial Gene Expression app finds marker genes with significantly different expression between different categories of cells, e.g. cell subtypes in a cell clustering or different cell branches in an infered cell differentiation. The app is based on the diff rank algorithm in the scanpy package.

  • Dimensionality Reduction (Autoencoder)

    This app computes a 2-dimensional non-linear embedding of cells based on their gene expression, representing local and global similarities. Based on autoencoder/ neural network implementation. Choose this app if you require a visualization of very large scRNA

  • Dimensionality Reduction (Parametric tSNE)

    This app computes a 2-dimensional non-linear embedding of cells, representing local similarities in their gene expression profiles. The app is based on a neural network implementation of the t-SNE loss function. Choose this app if you require a tSNE-like embedding of very large scRNA data sets (up to millions of cells).

  • Dimensionality Reduction (Truncated SVD)

    This app performs efficient linear dimensionality reduction on large, sparse matrices to a lower dimensionalty to enable further downstream processing with clustering or embedding tasks. Use this app as a preprocessing tool to reduce data set size for algorithms that typically do not work in very high-dimensional spaces like k-means.

  • Expression over Pseudotime Plot

    This app visualizes the expression of multiple genes over pseudo time in a line chart. The expression plot allows to select individual genes of interest.

  • Filtering (Cell Ids and/or Gene IDs)

    This app filters an expression matrix by cell and/or gene IDs. Supplied IDs can be used as blacklist (removing these entries from the expression matrix) or as whitelist (keeping only these entries in the expression matrix).

  • Filtering (Data Quality Check for Genes and Cells)

    This app filters an expression matrix by data quality statistics. It supports three types of thresholding filters: (i) gene detection thresholds for both cells and genes, (ii) minimum number of cells per gene or (iii) minimum number of genes per cell.

  • Filtering (Gene Lists)

    This app filters an expression matrix by a gene list (Entrez nomenclature). Supplied gene IDs can be used as blacklist (removing these entries from the expression matrix) or as whitelist (keeping only these entries in the expression matrix). For filtering mitochronidrial genes the app already provides a predefined gene list.

  • Filtering (Highly Variable Genes)

    This app filters an expression matrix to the most variable genes.

  • Functional Analysis (Fisher Test)

    The Functional Analysis App assess which biological processes are overrepresented in a set of genes (e.g. marker genes from different cell clusters) using Fisher’s exact test.

  • Functional Analysis (SEA)

    The functional analysis app assess which biological processes represented by GO terms, are overrepresented in a set of genes (e.g. marker genes from different cell clusters) using singular enrichment analyses. The app is based on a Fisher’s exact test.

  • Gene Type Chart

    The Gene Type bar chart visualizes the distribution of gene types among the genes in the current data set.

  • Heatmap

    The Heatmap app creates a standard heatmap plot. Use this app e.g. to visualize the average gene expression of DE genes or the p-values of biological processes in a cluster.

  • Hierarchical Clustering Report

    This graphical report provides a dendrogram and scatterplot for visualizing the results of the hierarchical clustering app. Use this app in combination with the according calculation app.

  • Image Display

    This app load images to display in FASTGenomics.

  • Pseudotime Analysis (Diffusion Map Based)

    The Pseudotime Analysis App embedds cells in 2-dimensional space and assigns them to a linage tree representing an inferred temporal development according to their gene expression. The algorithm is based on the scanpy package.

  • Sankey Diagram

    This app produces a sankey plot. Use this app e.g. to analyze predicted classes in the classification.

  • Table

    The Table app visualizes an arbitrary CSV file as a table. Use this app e.g. to report the statistics of all differently expressed genes.

  • Violin and Box Plot for Cluster Profiles

    This visualization creates a violin or box plot to display a distribution of counts in cell clusters after a cluster analysis.

Workflows

  • Subtype Discovery

    The Subtype Discovery workflow detects and explores clusters using parametric t-SNE, H-DBSCAN clustering, DE gene identification with GLMs and functional enrichment analysis using Fisher’s exact test and the GO database.

  • Subtype Discovery (LIMES Bonn)

    The Subtype Discovery (LIMES Bonn) workflow detects and explores clusters using the Seurat clustering, DE gene identification with GLMs and functional enrichment analysis using Fisher’s exact test and the GO database.

  • Pseudotime Analysis

    The Pseudotime Analysis reconstructs and explores the progression of dynamic biological processes using diffusion pseudotime from the scanpy package, DE gene identification using scanpy diffrank and functional enrichment analysis using Fisher’s exact test and the GO database.