Approximate distance correlation for selecting highly interrelated genes across datasets.

Area of Science:

• Genomics and Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Biological omics datasets are rapidly accumulating, necessitating methods to decode cross-dataset gene relationships.
• Existing methods struggle to identify interrelated genes across datasets with different samples.
• Quantifying gene associations across distinct biological datasets remains a challenge.

Purpose of the Study:

• To propose and validate Approximate Distance Correlation (ADC) for selecting statistically significant interrelated genes across two distinct biological datasets.
• To address the limitations of existing methods in detecting gene interrelationships across datasets with varying sample origins.
• To establish a scalable computational tool for cross-dataset gene association analysis.

Main Methods:

• Approximate Distance Correlation (ADC) identifies the k most correlated genes as approximate observations for each target gene.
• Distance Correlation (DC) is calculated for the target gene across two datasets using these approximate observations.
• The Benjamini-Hochberg adjustment is applied to control the false discovery rate for selected genes.

Main Results:

• ADC effectively selects highly interrelated genes across diverse datasets, validated through simulations and four real-world applications.
• Applications included cancer RNA-seq, mouse hematopoietic scRNA-seq, pancreatic islet scRNA-seq, and PBMC scATAC-seq/scRNA-seq data.
• The number of identified interrelated genes serves as a metric for dataset similarity, characterizing differences in cell types and technologies.

Conclusions:

• ADC is a powerful and scalable tool for uncovering biologically significant interrelated genes across different datasets.
• The method provides a robust approach to quantify gene associations even when datasets originate from different samples.
• ADC's ability to measure dataset similarity has implications for comparative genomics and understanding biological variation.