CogentDS FAQs

Cogent NGS Discovery Software (CogentDS) is a bioinformatics software for downstream processing and visualization of NGS data generated using Takara Bio NGS reagent kits and Illumina sequencing platforms, and processed using CogentAP.

CogentDS provides a guided workflow for end-to-end single-cell RNA (scRNA), bulk RNA, and single-cell DNA (scDNA) analysis.

CogentDS runs on a user workstation (laptop or desktop) and requires R. It is supported on:

• Windows 11
• MacOS Mojave (10.14) or higher
• Linux CentOS 6.9 or higher

For installing CogentDS, you need to install R version 4.4 or higher, as well as RStudio (in a version that is compatible with the R version).

It is advisable to uninstall the older version of CogentDS before installing the newer version to avoid version conflicts for the various required packages.

If there are enough computational resources, it is possible to run all three apps (scRNA, bulk RNA, and scDNA) simultaneously.

Input into CogentDS depends on the workflow module:

• For scRNA (analysis mode): Use CogentDS.analysis.rds, a .rds output from CogentAP, or the raw gene-count matrix and stats files
• For scRNA (discovery mode): Use CogentDS.analysis.rds, a .rds output file from CogentDS scRNA analysis mode
• For scRNA barcode rank plot: Use demultiplexed_fastqs_counts_all.estimated.csv from CogentAP dry run
• For scDNA: Use CogentDS_scDNA_analysis.rds, a .rds output from CogentAP
• For Bulk RNA: Use CogentDS.analysis.rds, a .rds output from CogentAP, or the raw gene-count matrix and stats files

CogentDS only accepts .rds files that match the structure created by CogentAP. Externally created files must conform to this format to be imported successfully.

After your analysis is complete, you can download a comprehensive HTML report. The report contains all the parameters and refinements applied during the analysis. In addition, all analyses can be downloaded as an .rds object. The .rds file that is created can be imported back into CogentDS for further visualization and analysis.

Pathway analysis, gene fusion, and clonotype overlays are not available in the report.

At any stage of the analysis, if you wish to stop, simply click on the “next” button until you reach the last page, where you can download the HTML report and the CogentDS-processed .rds file. Except for normalization and linear dimension reduction, which are required steps, all the other steps can be bypassed to reach the last stage of .rds file generation. This file can then be imported into CogentDS later to continue with the analysis.

Analysis mode allows you to take the output of CogentAP, perform QC and normalization, and run advanced analysis. This includes filtering genes based on the number of cells expressing them, filtering cells based on the number of expressed genes, ambient RNA correction, and QC steps. You can follow up with advanced investigations such as cell-type annotation, pathway enrichment, and immune, fusion, and clonotype analyses.

In Discovery mode, you can import a pre-analyzed CogentDS dataset and visualize the data using different plots, including ridge plots, dot plots, violin plots, and feature plots. The main purpose of Discovery mode is to allow you to visually inspect the features/genes or metadata fields.

Yes, analysis in CogentDS can be restricted to the barcodes of interest by using a metadata file during data import and selecting the samples you want to analyze.

CogentDS uses DecontX for ambient RNA removal. DecontX is a Bayesian method that estimates contamination levels in single-cell RNA-seq data and removes ambient RNA by modelling each cell’s expression profile as a mixture of native and contaminant transcripts. Rather than removing contaminated cells, DecontX deconvolves each expression profile into native and contaminant components, returning corrected counts for all input cells.

To proceed with uncorrected data, simply click on “Skip ambient RNA correction” and proceed with the next steps.

The “Pre-QC and QC analysis” page displays values for the “Maximum number of genes/features/cells”, the “Maximum intergenic reads percentage”, the “Maximum ribosomal reads percentage”, and the “Maximum percentage of mitochondrial genes”. The values are determined as follows:

• Maximum number of genes/features/cells: this is the 95th percentile of the number of features across all cells in the dataset
• Maximum intergenic reads percentage: this is calculated using the upper limit of the Mean Absolute Deviation (MAD) confidence interval + MAD
• Maximum ribosomal reads percentage: this is calculated using the upper limit of the MAD confidence interval + MAD
• Maximum percentage of mitochondrial genes: this is the 90th percentile of the mitochondrial gene percentage across all cells in the dataset

Data will not be renormalized. CogentDS picks raw counts instead of normalized counts. However, if the data have been corrected for ambient RNA, then the corrected data will be used when CogentDS imports the .rds file.

The counts filter is used for the differential expression (DE) step to filter out genes that do not meet a minimum number of counts. To pass through the counts filter, genes must have a minimum number of reads (i.e., the selected filter value) in a minimum number of samples (i.e., the number of samples in your smallest experimental group).

Here is an example:

• In your experiment, there are two experimental groups: Group A and Group B
• Group A has 8 samples and Group B has 4 samples; in this case, the number of samples in your smallest group is 4
• You select a counts filter value of 10
• The counts filter will remove genes that do not have at least 10 reads in at least 4 out of the 12 total samples

A metadata file is an optional input file that can be used to provide additional characteristics, such as sample biotype, to associate with the barcodes. These additional characteristics can be used for further data analysis.

While optional, a metadata file is necessary for a meaningful MA plot. An MA plot shows the relationship between fold change (M values) and average log expression across samples (A values) for differential expression analysis.

A metadata file is a .csv file with sequencing sample barcodes in one column and associated metadata in other columns.

The sample distance plot uses Euclidean distance to calculate and assess similarity between samples.

The blue dots are the statistically significant genes. The MA plot is generated using the DESeq2 function. For each gene, it plots the log2 fold change and the measure of normalized counts across all samples. It performs the Wald test, and genes with an adjusted p value of less than 0.1 are shown in blue.

While performing significance testing for the MA plot, DESeq2 (by default) chooses the first condition in the metadata file as the control, and uses that to perform pairwise comparisons for all of the other conditions.

DESeq2 is used for DE analysis. It uses the Wald test for pairwise comparison.

The Wald test is for pairwise comparison. Therefore, a multi-condition test can only be performed using individual pairwise comparisons.

The Benjamini-Hochberg procedure is used to obtain an adjusted p value in the DE analysis.

During data import, column-wise normalization is performed, where the mean of each column is computed, followed by division of each column value by its mean. Normalization is performed for the raw count file.

The size range for CN would be determined by the bin size selected for analysis in CogentAP and the minimum number of bins per segment. The current CogentAP pipeline supports 500 kb and 1 mb bins. For a copy number variation (CNV) to be detected, at least 5 consecutive bins are required.

CogentDS can detect amplifications and deletions. For amplification events, the CN is capped at 6. Homozygous and heterozygous deletion events with CNs of 0 and 1 are also detected.

Outliers are noisy barcodes that need to be excluded from analysis. Barcodes with Gini coefficients (uniformity of genome coverage) above the threshold are flagged as outliers. CogentDS uses the following calculation to define outliers: Q3 + (n × IQR), where Q3 is the third quartile, n is the outlier detection threshold set by the user, and IQR is the interquartile range. By default, n is set to 3.

Outlier detection is computed in all the modules and is based on the Gini coefficient.

CogentDS uses the natural log of normalized read counts in the Loess plot.

It is not possible to draw the heatmap for a single barcode. A single barcode is treated as a numeric vector in R, while a 2D matrix is required for the heatmap construction.

The CN profile window shows all barcodes in the dropdown view, regardless of any filtration in the previous step. Up to 3 barcodes can be selected at a time for viewing.

The "Predicted Ploidy" shown in the CN profile plot is the decimal scaling factor (multiplier) that best fits the normalized read depth to integer copy numbers. The "Ploidy" in the metadata is the median of the inferred integer copy numbers across the entire genome for each sample. It summarizes the most common copy number state.