Multitask learning-driven identification of novel antitrypanosomal compounds.

Area of Science:

• Drug discovery and development
• Computational chemistry
• Parasitology

Background:

• Chagas disease and human African trypanosomiasis cause significant global health burdens.
• There is an urgent need for new therapeutic agents to combat these neglected tropical diseases.

Purpose of the Study:

• To develop an explainable multitask AI pipeline for profiling compound activity against trypanosomes.
• To identify novel antitrypanosomal compounds through virtual screening.

Main Methods:

• Creation of an explainable multitask AI pipeline.
• Profiling compound activity against three key trypanosome species: *Trypanosoma brucei brucei*, *Trypanosoma brucei rhodesiense*, and *Trypanosoma cruzi*.
• Virtual screening of chemical compounds.

Main Results:

• The AI pipeline successfully identified four new experimental antitrypanosomal compounds: LC-3, LC-4, LC-6, and LC-15.
• Compound LC-6 demonstrated potent activity with IC50 values between 0.01-0.072 μM and high selectivity indices (>10,000).

Conclusions:

• The multitask protocol provides predictive and interpretable results for virtual screening of antitrypanosomal compounds.
• This approach has the potential to enhance hit discovery rates for Chagas disease and human African trypanosomiasis drug development projects.