Analyzing omics data based on sample network.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Accurate feature identification from complex omics data is crucial for disease diagnosis.
• Existing methods may not fully capture intricate relationships within molecular expression data.

Purpose of the Study:

• To propose a novel feature selection algorithm, Feature Selection based on Sample Network (FS-SN), for omics data analysis.
• To enhance the identification of significant features for disease diagnosis and biological mechanism discovery.

Main Methods:

• Constructing a sample network based on molecular expression-level neighbor relationships.
• Evaluating feature discriminating ability using sample network topology (intra-group vs. inter-group edges).
• Employing a gravitational interaction model to remove redundant features.

Main Results:

• FS-SN demonstrated superior performance compared to ERGS, mRMR, ReliefF, ATSD-DN, and INDEED on ten public omics datasets.
• The algorithm achieved higher accuracy, sensitivity, and specificity in most comparative cases.
• FS-SN effectively identified key features indicative of disease occurrence and development.

Conclusions:

• FS-SN is an effective algorithm for analyzing omics data by leveraging sample network topology.
• The method successfully identifies critical features and aids in understanding disease-related biological mechanisms.
• FS-SN offers a promising approach for advancing disease diagnosis studies using omics data.