diffGEK: differential gene expression kinetics

TL;DR

The paper introduces diffGEK, a new method to compare gene expression kinetics across biological conditions using single-cell RNA data.

Contribution

diffGEK allows smooth and trajectory-based estimation of transcriptional rates, overcoming limitations of previous models.

Findings

diffGEK identifies genes with altered transcription, splicing, or degradation rates in mutant versus wild type mice.

Compensatory changes in gene expression rates can mask dynamic changes in conventional expression analysis.

The method provides a robust pipeline for discovering mechanistic differences missed by traditional approaches.

Abstract

A defining characteristic of all metazoan organisms is the existence of different cell states or cell types, driven by changes in gene expression kinetics, principally transcription, splicing and degradation rates. The RNA velocity framework utilizes both spliced and unspliced reads in single cell mRNA preparations to predict future cellular states and estimate transcriptional kinetics. However, current models assume either constant kinetic rates, rates equal for all genes, or rates completely independent of progression through differentiation. Consequently, current models for rate estimation are either underparametrized or overparametrized.

Here, we developed a new method (diffGEK) which overcomes this issue, and allows comparison of transcriptional rates across different biological conditions. diffGEK assumes that rates can vary over a trajectory, but are smooth functions of the differentiation process. Analysing Jak2 V617F mutant versus wild type mice for erythropoiesis, and Ezh2 KO versus wild type mice in myelopoiesis, revealed which genes show altered transcription, splicing or degradation rates between different conditions. Moreover, we observed that, for some genes, compensatory changes between different rates can result in comparable overall mRNA levels, thereby masking highly dynamic changes in gene expression kinetics in conventional expression analysis. Collectively, we report a robust pipeline for comparative expression analysis based on altered transcriptional kinetics to discover mechanistic differences missed by conventional approaches, with broad applicability across any biomedical research question where single cell expression data are available for both wild type and treatment/mutant conditions.

This study does not include new data. All the codes are available on github: https://github.com/mebarile/transcriptional_kinetics.

Figures