GEARS Explained: How Graph Neural Networks Predict the Unseen

Fri, 05 Jun 2026 07:19:29 +0800

A deep dive into the model that can forecast what happens when you knock out genes nobody has ever touched.

The Problem

You’re a biologist. You have a Perturb-seq dataset — 100,000 cells, each with one or two genes CRISPR-knocked out, plus single-cell RNA-seq readouts of the full transcriptome. You’ve experimentally perturbed 100 single genes and 130 two-gene combos. But there are 4,950 untested pairwise combinations. Running them all would cost a fortune.

Learning and Practice of Single-Cell Sequencing

Mon, 09 Mar 2026 16:09:29 +0800

Abstract

Single-cell RNA sequencing (scRNA-seq) has become an essential technique for studying cellular heterogeneity and complex biological systems. During this winter research training, I systematically studied the general workflow and analytical principles of scRNA-seq based on Single Cell Best Practices and related resources, covering key steps such as data preprocessing, quality control, normalization, dimensionality reduction, clustering, and cell type annotation. To consolidate the knowledge, I first reproduced a complete analysis pipeline using publicly available immune cell data to familiarize myself with standard procedures and tools, then independently applied the same workflow to a publicly available human brain infection-related single-cell dataset, identifying distinct cell populations and infection-associated transcriptional changes across cell types. Overall, this training deepened my understanding of scRNA-seq data analysis, demonstrated the adaptability of standardized workflows to diverse biological contexts, and provided preliminary insights into cellular responses in infected human brain tissue as a foundation for further studies.

Single Cell on Mike's Blog

GEARS Explained: How Graph Neural Networks Predict the Unseen

The Problem

Learning and Practice of Single-Cell Sequencing

Abstract