Protocol to perform cell-type-specific transcriptome-wide association study using scPrediXcan framework

TL;DR

This paper introduces a protocol for cell-type-specific transcriptome-wide association studies using the scPrediXcan framework, which uses deep learning to predict gene expression from DNA and epigenetic data.

Contribution

A protocol for scalable, cell-type-specific TWAS using deep learning without specialized hardware.

Findings

scPrediXcan enables cell-type-specific TWAS using single-cell data and epigenetic features.

Models can be trained for various cell types without requiring specialized hardware.

The framework allows prioritization of causal genes across different cellular contexts.

Abstract

The scPrediXcan framework enables cell-type-specific transcriptome-wide association studies (TWASs) by integrating deep learning-based prediction of gene expression from DNA sequence and epigenetic features. We present a protocol for scPrediXcan: training cell-type-specific models for expression prediction, predicting personalized expression, and testing associations with genome-wide association study (GWAS) summary statistics. This framework produces scalable TWAS models for different cellular contexts with minimal computational burden.

For complete details on the use and execution of this protocol, please refer to Zhou et al.1

•Perform cell-type-specific TWASs with scPrediXcan using single-cell data•Train gene expression models for various cell types without specialized hardware•Prioritize causal genes for diseases across cellular contexts

Perform cell-type-specific TWASs with scPrediXcan using single-cell data

Train gene expression models for various cell types without specialized hardware

Prioritize causal genes for diseases across cellular contexts

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

The scPrediXcan framework enables cell-type-specific transcriptome-wide association studies (TWASs) by integrating deep learning-based prediction of gene expression from DNA sequence and epigenetic features. We present a protocol for scPrediXcan: training cell-type-specific models for expression prediction, predicting personalized expression, and testing associations with genome-wide association study (GWAS) summary statistics. This framework produces scalable TWAS models for different cellular contexts with minimal computational burden.

Figures