scMEDAL for the interpretable analysis of single-cell transcriptomics data with batch effect visualization using a deep mixed effects autoencoder | JoVE Visualize

Area of Science:

Computational Biology and Bioinformatics
Genomics and Transcriptomics
Single-cell Analysis

Background:

Single-cell RNA sequencing (scRNA-seq) offers insights into cellular heterogeneity but is confounded by technical and biological batch effects.
Current batch correction methods often discard or suppress these effects, limiting deeper analysis.
There is a need for methods that can quantify and model batch effects rather than merely remove them.

Purpose of the Study:

To introduce scMEDAL (single-cell Mixed-Effects Deep Autoencoder Learning), a novel framework for modeling batch effects in scRNA-seq data.
To separately model batch-invariant and batch-specific variations to improve data accuracy and interpretability.
To enable retrospective analyses and predictions of cellular expression across different batches.

Main Methods:

scMEDAL utilizes two complementary autoencoder networks: one for batch-invariant representation via adversarial learning, and a Bayesian autoencoder for batch-specific representation.
The framework models both fixed and random effects, allowing for detailed analysis of batch contributions.
Genomap projections are employed for retrospective analyses, predicting cellular expression in hypothetical batches.

Main Results:

scMEDAL successfully suppresses batch effects while retaining valuable batch-specific biological and technical variation.
Evaluations across diverse conditions (autism, leukemia, cardiovascular) and cell types confirm enhanced accuracy and interpretability.
The framework enables accurate predictions of disease status, donor group, and cell type by integrating batch-agnostic and batch-specific latent spaces.

Conclusions:

scMEDAL provides a powerful approach to disentangle technical and biological variations in scRNA-seq data.
The ability to model and predict batch-specific effects enhances the utility of scRNA-seq for deeper biological insights.
scMEDAL is a valuable tool for understanding data acquisition nuances and cellular heterogeneity.