Characterizing Human Cell Types and Tissue Origin Using the Benford Law.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Massive transcriptomic datasets present challenges for meaningful analysis.
• Novel interdisciplinary approaches are needed to extract biological insights.
• The Benford Law (BL) describes digit distribution in numerical datasets.

Purpose of the Study:

• To investigate if Benford Law adherence in gene expression data can differentiate cell types and tissue origins.
• To evaluate the accuracy of using Benford adherence scores for sample classification.
• To compare this method with traditional approaches like mean-expression and differential expression.

Main Methods:

• Analysis of large single-cell and bulk RNA-sequencing datasets.
• Calculation of Benford Law adherence scores for specific genes.
• Application of machine-learning algorithms using Benford adherence scores as features.
• Comparison of classification accuracy with mean-expression and differential expression methods.

Main Results:

• Genes exhibiting specific first-digit distributions successfully differentiated between cell types and tissue origins.
• The Benford Law-based feature selection method demonstrated high separation accuracy.
• This novel method outperformed the mean-expression level approach and matched the differential expression approach in accuracy.

Conclusions:

• The Benford Law provides a powerful tool for extracting biological insights from large-scale genomic data.
• This approach can be applied to diverse biomedical challenges, including sample origin determination and cancer subtype analysis.
• Benford Law adherence scores offer a simple yet effective feature for machine learning in genomics.