Leveraging Sex-Dependent Gene Expression for Deconvolution of Pooled snRNA-Seq Data

Strategic Insights -

A study published on biorxiv.org has demonstrated the potential of using sex-dependent gene expression patterns to deconvolute pooled single-nucleus RNA sequencing (snRNA-seq) data. This approach enables the incorporation of pooled barcoding and sequencing protocols, increasing data throughput and analytical sample size without requiring increases in experimental sample size and sequencing costs. The researchers trained machine learning models using previously published snRNA-seq data from the rat ventral tegmental area, achieving high accuracy in cell sex classification (93-95%) and outperforming simple classification models using only sex chromosome gene expression.

Key Takeaways:

  • The study demonstrates the feasibility of using sex-dependent gene expression patterns for the deconvolution of pooled snRNA-seq data.
  • Machine learning models trained on previously published snRNA-seq data from the rat ventral tegmental area achieved high accuracy in cell sex classification (93-95%).
  • The models outperformed simple classification models using only sex chromosome gene expression (88-90%).
  • The generalizability of the models was assessed using an additional published data set from the rat nucleus accumbens, with model performance remaining highly accurate in cell sex classification (90-92%).
  • The study provides a model for future snRNA-seq studies to perform sample deconvolution using a two-sex pooled sample sequencing design.
  • The benchmarking of machine learning approaches to deconvolve sample identification from inherent sample features is a significant contribution of the study.

Statistics:

  • The accuracy of cell sex classification using machine learning models trained on snRNA-seq data from the rat ventral tegmental area ranged from 93-95%.
  • The accuracy of cell sex classification using simple classification models using only sex chromosome gene expression ranged from 88-90%.
  • The generalizability of the models trained on the rat ventral tegmental area data to the rat nucleus accumbens data was assessed using an additional published data set.
  • The performance of the models on the rat nucleus accumbens data remained highly accurate in cell sex classification (90-92%).

Sources:

  • biorxiv.org/content/10.1101/2024.11.29.626066v2