Using Machine Learning to Predict Synergistic Antimalarial Compound Combinations With Novel Structures.

Area of Science:

• Computational chemistry
• Parasitology
• Drug discovery

Background:

• Plasmodium falciparum causes lethal malaria, necessitating new treatments due to rapid resistance.
• Discovering synergistic drug combinations is crucial but experimentally challenging.
• Existing methods for screening drug combinations are not exhaustive.

Purpose of the Study:

• To develop and validate a machine learning model, Combination Synergy Estimation (CoSynE), for predicting synergistic antimalarial drug combinations.
• To expedite the discovery of novel drug combinations to combat antimalarial resistance.
• To assess the efficiency of CoSynE in reducing experimental screening efforts.

Main Methods:

• Applied CoSynE, a machine learning approach, to a dataset of 1,540 antimalarial drug combinations.
• Utilized prior experimental screening data and compound molecular structures as input for the model.
• Performed cross-validation and prospective validation with novel compounds.

Main Results:

• CoSynE showed significant enrichment of synergistic predictions compared to random selection (up to 2.74x in cross-validation).
• Prospective validation confirmed 45% of predicted synergistic combinations (9 out of 20), achieving a 1.70x enrichment.
• The predictive approach reduced experimental effort by 41% and identified novel synergistic combinations, often involving efflux inhibitors.

Conclusions:

• Predictive modeling, specifically CoSynE, can significantly accelerate the discovery of synergistic antimalarial drug combinations.
• Synergy in drug combinations is not random and can be predicted using machine learning.
• The CoSynE approach is broadly applicable for discovering novel drug combinations in other therapeutic areas.