Variable-selection ANOVA Simultaneous Component Analysis (VASCA).

Area of Science:

• Multivariate data analysis
• Bioinformatics
• Omics data analysis

Background:

• ANOVA Simultaneous Component Analysis (ASCA) is widely used for designed experiments but struggles with high-dimensional omics data.
• The holistic testing approach of ASCA can miss significant effects present in only a few variables (biomarkers).
• Existing methods often overlook subtle yet important biological signals in large omics datasets.

Purpose of the Study:

• To introduce Variable-selection ASCA (VASCA), a novel method generalizing ASCA with variable selection capabilities.
• To enhance the statistical power of ASCA for detecting effects in high-dimensional data.
• To maintain low Type-I error rates while improving the identification of biomarkers.

Main Methods:

• Developed VASCA by integrating variable selection into the ASCA framework.
• Evaluated VASCA using extensive simulations and a real-world multi-omic clinical dataset.
• Compared VASCA's performance against standard ASCA and false discovery rate controlling procedures.

Main Results:

• VASCA demonstrated superior statistical power compared to ASCA and common false discovery rate methods.
• The method effectively identifies significant effects encoded by a small subset of variables.
• VASCA proved useful for exploratory data analysis, outperforming Partial Least Squares Discriminant Analysis (PLSDA) and its sparse variant.

Conclusions:

• VASCA offers a powerful and reliable approach for analyzing high-dimensional omics data from designed experiments.
• The method successfully addresses the limitations of traditional ASCA in detecting subtle, biomarker-driven effects.
• VASCA provides a valuable tool for biomarker discovery and interpretation in complex biological studies.