Disease characterization using a partial correlation-based sample-specific network.

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Single-sample networks (SSNs) offer insights into individual disease mechanisms and personalized medicine.
• Existing SSN methods may not effectively distinguish direct from indirect molecular interactions.

Purpose of the Study:

• To develop a computational method, partial correlation-based single-sample network (P-SSN), for inferring direct interactions from single-sample data.
• To validate P-SSN's efficacy in analyzing complex biological datasets, including tumor and single-cell data.

Main Methods:

• Developed P-SSN to infer networks from single-sample data using a reference dataset.
• P-SSN specifically retains direct interactions by excluding indirect ones.
• Applied P-SSN to The Cancer Genome Atlas tumor data and single-cell data.

Main Results:

• P-SSN effectively predicts driver mutation genes (DMGs) from single-sample data.
• The method successfully identifies disease subtypes and classifies single cells.
• Network distance, derived from P-SSN, aids in complex disease subtyping and single-cell clustering.

Conclusions:

• P-SSN is a robust computational tool for constructing accurate single-sample networks.
• The method enhances the prediction of critical genes and improves the classification of complex biological samples.
• P-SSN facilitates advancements in personalized medicine and understanding disease heterogeneity.