SLIDE: Significant Latent Factor Interaction Discovery and Exploration across biological domains

Area of Science:

• Computational Biology
• Machine Learning
• Genomics

Background:

• High-dimensional omic datasets from modern technologies present analytical challenges due to data modality differences, multicollinearity, and irrelevant features.
• Integrating and analyzing these complex datasets is crucial for understanding biological systems but remains difficult with current methods.

Purpose of the Study:

• To introduce Significant Latent Factor Interaction Discovery and Exploration (SLIDE), an interpretable machine learning technique for identifying interacting latent factors in high-dimensional omic data.
• To provide a method with theoretical guarantees for latent factor identifiability and inference, coupled with rigorous false discovery rate control.

Main Methods:

• Development of SLIDE, a novel interpretable machine learning technique.
• Application of SLIDE to single-cell and spatial omic datasets.
• Comparison of SLIDE's performance against state-of-the-art approaches, including other latent factor methods.

Main Results:

• SLIDE successfully identified significant interacting latent factors underlying diverse molecular, cellular, and organismal phenotypes.
• The method demonstrated comparable or superior performance to existing state-of-the-art approaches.
• SLIDE provided biological inference capabilities beyond mere prediction, offering deeper insights.

Conclusions:

• SLIDE is a versatile and powerful engine for biological discovery from complex multiomic datasets.
• The technique offers robust analysis with theoretical guarantees and effective false discovery rate control.
• SLIDE advances the integration and interpretation of high-dimensional omic data for biological research.