Embracing the informative missingness and silent gene in analyzing biologically diverse samples

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Bioinformatics tools are crucial for identifying molecular features distinguishing phenotypic groups.
• Common tools struggle with biologically diverse samples, high missing data rates, and complex group comparisons.
• Accurate analysis of missing data and gene signatures is vital for biological interpretation.

Purpose of the Study:

• To develop a bioinformatics software suite (ABDS) for analyzing biologically diverse samples.
• To address challenges in missing value imputation, signature gene detection, and multi-group visualization.
• To enhance the accurate detection of interpretable molecular signals in complex biological data.

Main Methods:

• Developed a mechanism-integrated group-wise pre-imputation scheme to preserve informative missingness.
• Extended a cosine-based one-sample test for detecting group-silenced signature genes.
• Designed a unified heatmap for parallel visualization of multiple sample groups.

Main Results:

• The ABDS tool suite effectively handles informative missingness and detects group-silenced signature genes.
• Comparative evaluations and biomedical showcases demonstrate the suite's effectiveness.
• The unified heatmap provides clear visualization for multiple sample groups.

Conclusions:

• The ABDS tool suite is an open-source R package that complements existing bioinformatics tools.
• It enables biologists to more accurately detect molecular signals in phenotypically diverse sample groups.
• ABDS facilitates robust analysis of complex biological datasets with missing information.