Predicting protein secondary structure is essential for understanding their function and accelerating drug discovery. A new study presents MOGP-MMF, a multi-objective genetic programming framework that addresses this challenge as an automated optimization problem.

MOGP-MMF Architecture

MOGP-MMF introduces a multi-view multi-level representation strategy, integrating evolutionary, semantic, and structural views to capture the protein folding logic. The framework evolves both linear and nonlinear fusion functions, capturing high-order feature interactions and reducing fusion complexity.

Optimization and Results

An improved multi-objective GP algorithm incorporates a knowledge transfer mechanism, utilizing prior evolutionary experience to guide the population toward global optima. Extensive tests on seven benchmark datasets demonstrate that MOGP-MMF surpasses state-of-the-art methods, particularly in Q8 accuracy and structural integrity. The code is open source.