[Paper Review] LightCPPgen: An Explainable Machine Learning Pipeline for Rational Design of Cell Penetrating Peptides

Gabriele Maroni, Filip Stojceski+4 more|arXiv (Cornell University)|May 31, 2024

RNA Interference and Gene DeliveryBiochemistry, Genetics and Molecular Biology3 citations

TL;DR

LightCPPgen introduces an explainable machine learning pipeline that combines a LightGBM-based predictive model with a genetic algorithm to enable rational, efficient, and interpretable de novo design of cell-penetrating peptides (CPPs). By leveraging 20 explainable features and optimizing for both penetrability and similarity to parent sequences, the framework reduces experimental burden by prioritizing high-potential candidates with minimal wet-lab screening.

ABSTRACT

Cell-penetrating peptides (CPPs) are powerful vectors for the intracellular delivery of a diverse array of therapeutic molecules. Despite their potential, the rational design of CPPs remains a challenging task that often requires extensive experimental efforts and iterations. In this study, we introduce an innovative approach for the de novo design of CPPs, leveraging the strengths of machine learning (ML) and optimization algorithms. Our strategy, named LightCPPgen, integrates a LightGBM-based predictive model with a genetic algorithm (GA), enabling the systematic generation and optimization of CPP sequences. At the core of our methodology is the development of an accurate, efficient, and interpretable predictive model, which utilizes 20 explainable features to shed light on the critical factors influencing CPP translocation capacity. The CPP predictive model works synergistically with an optimization algorithm, which is tuned to enhance computational efficiency while maintaining optimization performance. The GA solutions specifically target the candidate sequences' penetrability score, while trying to maximize similarity with the original non-penetrating peptide in order to retain its original biological and physicochemical properties. By prioritizing the synthesis of only the most promising CPP candidates, LightCPPgen can drastically reduce the time and cost associated with wet lab experiments. In summary, our research makes a substantial contribution to the field of CPP design, offering a robust framework that combines ML and optimization techniques to facilitate the rational design of penetrating peptides, by enhancing the explainability and interpretability of the design process.

Motivation & Objective

• To address the challenge of rational, efficient, and interpretable de novo design of cell-penetrating peptides (CPPs) with minimal experimental iteration.
• To develop a predictive model that identifies key physicochemical and structural features influencing CPP translocation capacity.
• To integrate machine learning with optimization algorithms to generate high-performing CPP sequences while preserving the original peptide's biological and physicochemical properties.
• To reduce the time and cost of wet-lab validation by prioritizing only the most promising candidates for synthesis.
• To enhance transparency in CPP design through feature importance analysis and model interpretability.

Proposed method

• A LightGBM-based predictive model is trained on 20 explainable features derived from peptide sequences to predict translocation capacity.
• The model is integrated into a genetic algorithm (GA) framework that optimizes for high penetrability scores while maximizing sequence similarity to the original non-penetrating peptide.
• The optimization process balances two objectives: maximizing predicted penetrability and preserving key physicochemical and biological properties of the parent sequence.
• Feature importance analysis is performed to interpret model predictions and identify critical determinants of CPP function.
• The pipeline generates candidate sequences in silico, focusing only on those with high predicted performance to minimize experimental screening.
• The framework is computationally efficient, enabling rapid exploration of the sequence space without exhaustive enumeration.

Experimental results

Research questions

• RQ1 Which 20 explainable features most strongly influence the translocation capacity of cell-penetrating peptides?
• RQ2 Can a hybrid machine learning and optimization pipeline effectively generate novel CPP sequences with high predicted performance?
• RQ3 To what extent can the design process preserve the original peptide's physicochemical and biological properties while enhancing penetrability?
• RQ4 How does the integration of model interpretability improve the rationality and transparency of CPP design?
• RQ5 Can the pipeline significantly reduce the number of required wet-lab experiments by prioritizing only the most promising candidates?

Key findings

• The LightGBM model achieved high predictive accuracy in estimating CPP translocation capacity using only 20 explainable features.
• Feature importance analysis revealed that charge distribution, hydrophobicity, and secondary structure propensity are key determinants of CPP efficiency.
• The genetic algorithm successfully generated novel peptide sequences with significantly higher predicted penetrability scores than the original non-penetrating peptide.
• The optimized sequences maintained high similarity to the parent sequence in terms of key physicochemical properties, ensuring retained stability and biological compatibility.
• The pipeline reduced the number of candidate sequences requiring experimental validation by focusing only on top-performing in silico candidates.
• The integration of explainability into the design process enables researchers to understand and validate the rationale behind sequence selection, enhancing trust and reproducibility.