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

Area of Science:

• Neuroscience
• Genomics
• Computational Biology

Background:

• Cell deconvolution methods are crucial for analyzing bulk tissue transcriptomics.
• Their performance and applications in human brain data are underexplored.
• Accurate cell-type proportion and gene expression estimation is vital for understanding brain complexity.

Purpose of the Study:

• To evaluate and compare nine deconvolution methods using diverse human brain and organoid datasets.
• To identify optimal methods for estimating cell proportions and gene expression from bulk tissue.
• To explore the biological applications of deconvoluted data in brain disorders.

Main Methods:

• Evaluated nine deconvolution methods using matched bulk tissue RNA sequencing (RNA-seq), single-cell/nuclei (sc/sn) RNA-seq, and immunohistochemistry data.
• Utilized transcriptomic data from 149 adult postmortem brains and 72 organoid samples (1,130,767 nuclei).
• Identified cell-type specific expression quantitative trait loci (eQTLs) using deconvoluted gene expression data.

Main Results:

• Dtangle demonstrated superior performance for estimating cell-type proportions.
• bMIND showed the best performance for estimating sample-wise cell-type gene expressions.
• Identified 25,273 cell-type eQTLs (decon-eQTLs) for eight brain cell types.
• Decon-eQTLs explained more schizophrenia GWAS heritability compared to bulk or single-cell eQTLs alone.
• Differential gene expression analysis revealed associations with Alzheimer's disease, schizophrenia, and brain development.

Conclusions:

• Dtangle and bMIND are recommended for cell deconvolution in human brain studies.
• Deconvoluted data provides valuable insights into the genetic architecture of brain disorders.
• The findings support the utility of deconvolution methods for analyzing complex human brain transcriptomic data and understanding disease mechanisms.