COTAN: scRNA-seq data analysis based on gene co-expression

Abstract

Estimating the co-expression of cell identity factors in single-cell is crucial. Due to the low efficiency of scRNA-seq methodologies, sensitive computational approaches are critical to accurately infer transcription profiles in a cell population. We introduce COTAN, a statistical and computational method, to analyze the co-expression of gene pairs at single cell level, providing the foundation for single-cell gene interactome analysis. The basic idea is studying the zero UMI counts’ distribution instead of focusing on positive counts; this is done with a generalized contingency tables framework. COTAN can assess the correlated or anti-correlated expression of gene pairs, providing a new correlation index with an approximate p-value for the associated test of independence. COTAN can evaluate whether single genes are differentially expressed, scoring them with a newly defined global differentiation index. Similarly to correlation network analysis, it provides ways to plot and cluster genes according to their co-expression pattern with other genes, effectively helping the study of gene interactions, becoming a new tool to identify cell-identity markers. We assayed COTAN on two neural development datasets with very promising results. COTAN is an R package that complements the traditional single cell RNA-seq analysis and it is available at https://github.com/seriph78/COTAN.

A multifunctional locus controls motor neuron differentiation through short and long noncoding RNAs

“…For each sample, cells were clustered by Seurat V3.2 (Stuart et al , 2019 ). Cluster uniformity was then checked using COTAN (Galfrè et al , 2021 ) evaluating whether < 1% of genes were over the threshold of 1.5 of GDI. If a cluster result was not uniform, with more than 1% of genes above 1.5, a new round of clustering was performed.…”

DeepVelo : Single-cell transcriptomic deep velocity field learning with neural ordinary differential equations

“…scRNA-seq experiments are particularly useful in dissecting these networks because of the single-cell resolution and cell type specificity. However, it is computationally challenging to detect strong correlations between genes using scRNA-seq data due to the noisy and sparse transcript measurements in single-cell experiments (40)(41)(42).…”

DeepVelo: Single-cell Transcriptomic Deep Velocity Field Learning with Neural Ordinary Differential Equations

“…This gene has also been previous identified by RNA velocity and experimentally validated as being necessary for granule cell formation; moreover, the deletion of Prox1 leads to the adoption of the pyramidal neuron fate (19). In addition, scDVF identified the top pyramidal neuron developmental driver gene as Runx1t1, which was recently shown to induce pyramidal neuron formation, with its deletion resulting in reduced neuron differentiation in vitro Due to sparse and noisy measurements, it is often challenging to detect strong correlation between genes in scRNA-seq, thereby making it difficult to find coherent functional modules in gene co-expression networks (22)(23)(24). However, denoising VAEs in scDVF can reduce the variability along a developmental trajectory due to the sparsity and noise associated with scRNA-seq (Fig.…”

“…4d). When we allowed a set of Due to sparse and noisy measurements, it is often challenging to detect strong correlations between genes in scRNA-seq, thereby making it difficult to find coherent functional modules in gene co-expression networks (38)(39)(40). However, denoising VAEs in DeepVelo can reduce the variability along a developmental trajectory caused by the sparsity and noise associated with scRNA-seq (Fig.…”