In-Silico Pipelined Medical Compound Generation for Neglected Tropical Diseases

Neglected Tropical Diseases (NTDs) continue to pose significant health burdens in impoverished regions worldwide. Despite their impact, pharmaceutical investment in developing treatments for NTDs remains limited, largely due to the resource-intensive nature of traditional drug discovery methods. The framework employs Generative Adversarial Networks (GANs) integrated with Long Short-Term Memory (LSTM)-based Recurrent Neural Networks (RNNs) to generate novel drug compounds. These compounds are then subjected to a comprehensive evaluation process, including screening through Lipinski's Rule of Five for oral bioavailability, Tanimoto Similarity analysis for structural comparison with existing NTD drugs, and Quantitative Structure-Activity Relationship (QSAR) modeling to predict medical potency. The results demonstrated that the proposed GAN-LSTM model successfully generated 100,000 novel compounds, of which 84.45% were identified as valid molecules based on Lipinski's Rule of Five. The generated compounds exhibited Tanimoto similarity scores ranging from 0.1448 to 0.75, with an average of 0.5102 for valid compounds, indicating a balance between novelty and similarity to existing NTD drugs. To validate the efficacy of the approach, the researchers compared it to a benchmark model using only LSTM-based RNNs, which generated only 56.46% valid molecules from the same number of samples. This comparison further underscores the potential of the GAN-enhanced generative AI approach to substantially expedite the drug discovery pipeline for NTD treatments, offering a promising pathway to address these long-neglected health challenges more efficiently and effectively.