Accurate sample deconvolution of pooled snRNA-seq using sex-dependent gene expression patterns.

Area of Science:

• Genomics
• Computational Biology
• Neuroscience

Background:

• Single nucleus RNA sequencing (snRNA-seq) provides high-resolution gene expression data.
• Current snRNA-seq methods often require pooling samples, losing individual sample data and increasing costs.
• Developing methods to deconvolve pooled data is crucial for maximizing throughput and analytical power.

Purpose of the Study:

• To demonstrate that sex-dependent gene expression patterns can be used to deconvolve pooled snRNA-seq data.
• To train and evaluate machine learning models for cell sex classification in pooled samples.
• To assess the generalizability of these models across different brain regions.

Main Methods:

• Utilized previously published snRNA-seq datasets from rat ventral tegmental area and nucleus accumbens.
• Trained machine learning models to classify cell sex based on differentially expressed genes between male and female rats.
• Compared performance of models using sex-dependent genes versus only sex chromosome genes.

Main Results:

• Machine learning models accurately predicted cell sex (90-92% accuracy) using sex-dependent gene expression.
• These models significantly outperformed those using only sex chromosome gene expression (69-89% accuracy).
• Models demonstrated high accuracy (89-90%) and generalizability to a different brain region (nucleus accumbens) without re-training.

Conclusions:

• Sex-dependent gene expression is a viable feature for deconvolving pooled snRNA-seq data.
• Machine learning approaches can effectively identify cell sex, enabling sample deconvolution.
• This strategy supports cost-effective, high-throughput snRNA-seq studies using pooled samples.