LengthLogD: A Length-Stratified Ensemble Framework for Enhanced Peptide Lipophilicity Prediction via Multi-Scale Feature Integration

TL;DR

This work tackles the challenge of predicting peptide lipophilicity ($R^2$) across molecules of varying length by introducing LengthLogD, a length-stratified framework that integrates multi-scale features from Morgan fingerprints, MACCS keys, RDKit/MOE descriptors, and graph topology. It employs category-specific ensembles with adaptive weighting to tailor predictions for short, medium, and long peptides, achieving $R^2$ values of 0.855, 0.816, and 0.882 respectively, and up to $R^2=0.891$ for long peptides. Ablation studies show length stratification (41.2% contribution) and topological features (28.5%) are key drivers, while MOE descriptors are indispensable for linear models; LengthLogD outperforms state-of-the-art peptide logD methods and reduces computational costs relative to MD-based approaches. The framework provides a practical, generalizable tool for peptide lead optimization and suggests pathways for extending to 3D descriptors and broader ADMET tasks.

Abstract

Peptide compounds demonstrate considerable potential as therapeutic agents due to their high target affinity and low toxicity, yet their drug development is constrained by their low membrane permeability. Molecular weight and peptide length have significant effects on the logD of peptides, which in turn influences their ability to cross biological membranes. However, accurate prediction of peptide logD remains challenging due to the complex interplay between sequence, structure, and ionization states. This study introduces LengthLogD, a predictive framework that establishes specialized models through molecular length stratification while innovatively integrating multi-scale molecular representations. We constructed feature spaces across three hierarchical levels: atomic (10 molecular descriptors), structural (1024-bit Morgan fingerprints), and topological (3 graph-based features including Wiener index), optimized through stratified ensemble learning. An adaptive weight allocation mechanism specifically developed for long peptides significantly enhances model generalizability. Experimental results demonstrate superior performance across all categories: short peptides (R^2=0.855), medium peptides (R^2=0.816), and long peptides (R^2=0.882), with a 34.7% reduction in prediction error for long peptides compared to conventional single-model approaches. Ablation studies confirm: 1) The length-stratified strategy contributes 41.2% to performance improvement; 2) Topological features account for 28.5% of predictive importance. Compared to state-of-the-art models, our method maintains short peptide prediction accuracy while achieving a 25.7% increase in the coefficient of determination (R^2) for long peptides. This research provides a precise logD prediction tool for peptide drug development, particularly demonstrating unique value in optimizing long peptide lead compounds.

Figures (7)

• Figure 1: Distribution of experimental logD values.
• Figure 2: Distribution of SMILES length representing the molecular size and complexity.
• Figure 3: Overall workflow of the LengthLogD framework.
• Figure 4: Scatter plots of true vs. predicted logD values for peptides in different length categories. From left to right: short, medium, and long peptides. The red dashed line indicates the ideal prediction line.
• Figure 5: Bar plot of $R^2$ scores under different ablation settings. MOE features contribute most significantly to the overall performance, especially for LR.