IPF-LASSO: Integrative L₁-Penalized Regression with Penalty Factors for Prediction Based on Multi-Omics Data

Area of Science:

• Biostatistics
• Bioinformatics
• Genomics

Background:

• Modern biotechnology enables collecting diverse high-dimensional molecular data (omics) from patient cohorts.
• Integrating multiple omics data types is crucial for accurate clinical outcome prediction and personalized medicine.
• Statistical methods for multi-omics integration and feature selection remain underdeveloped.

Purpose of the Study:

• To propose a novel penalized regression method for integrating multiple omics data modalities.
• To perform feature selection and predict clinical outcomes using integrated omics data.
• To develop an R package (ipflasso) for reproducible research.

Main Methods:

• A penalized regression approach assigning distinct penalty factors to different omics data modalities.
• Feature selection and prediction using the proposed Integrative LASSO with Penalty Factors (IPF-LASSO).
• Data-driven selection of penalty factors via cross-validation or practical considerations.

Main Results:

• IPF-LASSO demonstrated competitive prediction performance in simulation studies.
• The method was compared against standard LASSO and sparse group LASSO.
• The approach was successfully applied to two real-life cancer datasets.

Conclusions:

• IPF-LASSO offers an effective strategy for integrating multiple omics data for prediction.
• The method facilitates feature selection critical for personalized medicine.
• The R package ensures reproducibility and practical application of the proposed method.