Improve UMI Normalization that Determines the Downstream Interpretation of Single-Cell Transcriptomes

To be composed soon after paper is published.

Below is only a piece of the project for now.

Contents

1. 🧬 Dissect sctransform()📝Blog | 🐙Git

The vst() function applies variance stabilizing transformation to UMI count data using a regularized Negative Binomial regression model. This removes unwanted effects from UMI data and returns Pearson residuals.

The original scripts have overly complex logic flow. This blog breaks down the function into clear sections, documenting each parameter, its possible values, and upstream decision logic. Flowcharts illustrate the function’s internal logic for easy tracing. Below dissects vst() implementation which is the core of sctranform().