Evaluating performance and applications of sample-wise cell deconvolution methods on human brain transcriptomic data.

Area of Science:

• Neuroscience
• Genomics
• Bioinformatics

Background:

• Cell deconvolution methods estimate cell proportions and gene expression in bulk tissues but require evaluation on human brain transcriptomic data.
• Understanding cell-type-specific gene expression is crucial for interpreting complex biological processes and disease mechanisms.

Approach:

• Nine deconvolution algorithms were assessed using matched bulk-tissue RNAseq, single-cell/nuclei (sc/sn) RNAseq, and immunohistochemistry data from 149 adult postmortem brains and 72 organoid samples.
• Performance was evaluated for estimating cell proportions and sample-wise cell-type gene expression.
• Cell-type specific expression quantitative trait loci (eQTLs) were identified using deconvoluted data.

Key Points:

• dtangle demonstrated superior performance for cell proportion estimation, while bMIND excelled in sample-wise cell-type gene expression estimation.
• 25,273 cell-type eQTLs (decon-eQTLs) were identified across eight brain cell types.
• Decon-eQTLs explained a greater proportion of schizophrenia GWAS heritability compared to bulk-tissue or single-cell eQTLs alone.

Conclusions:

• The study validates and compares deconvolution methods for human brain transcriptomics, highlighting dtangle and bMIND.
• Deconvoluted data provides valuable insights into cell-type-specific gene regulation and its contribution to complex traits like schizophrenia.
• The findings offer a robust framework for applying deconvolution in neuroscience research and disease association studies.