# Multi-feature fusion network with marginal focal dice loss for multi-label therapeutic peptide prediction

## TL;DR

This paper introduces a new deep learning model for predicting the functions of multi-functional therapeutic peptides using feature fusion and a novel loss function to handle class imbalance.

## Contribution

The paper proposes a novel multi-feature fusion network with a marginal focal dice loss for improved multi-label therapeutic peptide prediction.

## Key findings

- The MFTP_MFFP model outperforms existing methods in all evaluation metrics.
- The gated feature fusion module enhances the integration of diverse peptide features.
- The proposed MFDL loss function effectively addresses class imbalance issues.

## Abstract

Accurately predicting the functions of multi-functional therapeutic peptides is crucial for the development of related drugs. However, existing peptide function prediction methods largely rely on either a single type of feature or a single model architecture, limiting prediction accuracy and applicability. Additionally, training better-performing models on datasets with class imbalance issues remains a significant challenge. In this study, we propose the multi-functional therapeutic peptide of multi-feature fusion prediction (MFTP_MFFP) model, a novel method for predicting the functionality of multi-functional therapeutic peptides. This approach uses various encoding techniques to process peptide sequence data, generating multiple features that help the model learn hidden information within the sequences. To maximize the effectiveness of these features, we propose a gated feature fusion module that efficiently integrates them. The module assigns learnable gating weights to each feature, optimizing integration and enhancing fusion efficiency. The fused features are then passed into a neural network model for feature extraction. Additionally, we propose a marginal focal dice loss function (MFDL) to address the class imbalance and improve the model’s prediction performance. Experimental results show that the MFTP_MFFP model outperforms existing models in all evaluation metrics, demonstrating its robustness and effectiveness in multi-functional therapeutic peptide prediction tasks.

Understanding the biological complexity and sequence ambiguity of multi-functional therapeutic peptides (MFTPs) is crucial for developing robust prediction models. In this work, we present MFTP_MFFP, a unified deep learning framework that combines biological domain knowledge with adaptive architectural strategies. Unlike prior methods that rely on fixed structures or single-source encodings, our approach integrates diverse biological features—including statistical, physicochemical, and evolutionary encodings—processed through fuzzy transformation to enhance robustness against uncertainty. By incorporating graph-based modeling and gated feature fusion, the framework captures both sequential and topological signals. Rather than deepening the network blindly, we emphasize structural adaptability and biological relevance. We hope this work offers insights into feature fusion, imbalance-aware loss design, and evolutionary optimization for future peptide-based prediction tasks.