Introduction and Background of the Project

Introduction

Schistosomiasis, caused by infection with schistosome parasites (e.g. Schistosoma mansoni), is one of the most important neglected tropical diseases that negatively impacts the bottom billion of the world’s poorest populations. Current treatment relies on a single chemotherapeutic agent, praziquantel (PZQ).  PZQ resistance has recently been found in one of the largest foci of schistosome-infected African populations undergoing long-term mass drug administration and highlights an urgent unmet medical need to identify new, PZQ-replacement, anti-schistosomal drugs.  Artificial intelligence (AI) offers transformative potential in drug discovery for schistosomiasis by addressing the challenges of cost, time and resource constraints.

AI-driven approaches, including generative models, can predict repurposable (or create novel) chemical structures active against schistosomes. Historically, large phenotypic screens against schistosomes are slow, resource intensive and attrition rates are high. Machine learning models have the potential to substantially increase the hit rate in phenotypic screens against schistosomes. If successfully optimized, this approach not only will reduce screening cost and use of animals but, very importantly, help to prime the pipeline for much needed starting points for new treatments for schistosomiasis.

Additionally, by mining genomic, proteomic, and chemical databases, AI can identify regions of schistosome targets and design molecules optimized for these targets (when compared to host homologs). Machine learning models can also predict the biological activity, toxicity, and pharmacokinetics of compounds, reducing the need for extensive and costly laboratory experiments.

When combined with existing high-throughput phenotyping assays and medicinal chemistry support, AI bridges critical gaps in funding and resource availability by offering cost-effective, scalable and rapid solutions to advance drug discovery for schistosomiasis, ultimately contributing to improved global health outcomes for underserved populations.

Project objective

The primary goal of this project is to deliver a small number of multi-lifecycle stage (schistosomula, juveniles and adults) active chemical series ready to enter the hit-to-lead phase of anti-schistosomal drug discovery.  To achieve this goal, we will combine AI-guided approaches for drug discovery with high-throughput, whole-organism imaging platforms to phenotype schistosomes and expertise in early-stage drug discovery for schistosomiasis including medicinal chemistry and DMPK.  The compounds developed here will be carried forward in subsequent projects as putative replacement drugs for praziquantel.

Project design

Using existing phenotypic datasets, AI models will first be built to identify chemical features associated with anti-schistosomal activity and run these trained models on large chemical libraries to identify new compounds predicted to display potential anti-schistosomal activity.

Using high-throughput phenotyping platforms, approximately 100 of the prioritized compounds will first be co-cultured with S. mansoni schistosomula (10 µM final concentration) for 72 h.  Active molecules will be subjected to dose response titrations and those displaying EC₅₀s less than 10 µM will subsequently be tested against 7-week adult male and female worms (20 µM final concentration) for 72 h. Actives will subsequently be tested in dose response titrations for EC₅₀ determinations.  Compounds with adult worm EC₅₀s less than 20 µM will subsequently be tested against 3-week juvenile worms in dose response assays.  Compounds demonstrating EC₅₀s less than 10 µM in schistosomula, 20 µM in adult worms and 20 µM juvenile worms will be progressed as hits. 

Confirmed hits will be tested for cytotoxicity, metabolic stability, aqueous solubility and lipophilicity. Focused hit expansion plans for the most promising actives will aim to assess the potential to improve potency and DMPK profile, explore synthetic tractability and identify key areas for further optimization.  Commercial analogs or, if available, analogs in our existing compound collection will also be tested. Series that meet Compound Progression Criteria for a Validated Hit Series will be screened against S. mansoni parasites during ex vivo co-culture. Compounds demonstrating EC₅₀s less than 3 µM in schistosomula, 3 µM in adult worms and 3 µM juvenile worms will be identified as H2L candidates.

These H2L candidates will form the basis of a future funding application to GHIT.

How can your partnership (project) address global health challenges?

Schistosomiasis is one of humanity’s most recalcitrant neglected tropical diseases.  It occurs in tropical and sub-tropical areas, kills thousands of individuals every year and contributes to an annual loss of up to 4.5 million disability adjusted life years in endemic communities.  Until a vaccine is registered for use, the cornerstone of schistosomiasis control remains mono-chemotherapy with praziquantel. This strategy’s sustainability is at risk due to the concern of praziquantel insensitive/resistant schistosomes developing and becoming widespread. Because of this threat, the revised 2021-2030 WHO roadmap for achieving schistosomiasis elimination includes actions to support new drug development. Building upon our expertise in computer science, parasitology and drug discovery, we will drive new approaches to identify urgently-needed, next-generation, anti-schistosomal candidates.

What sort of innovation are you bringing in your project?

We will create novel AI-driven pipelines to identify potent, anti-schistosomal compounds for validation in high-throughput ex vivo assays.  Medicinal chemistry optimization of these compounds will subsequently be pursued and the efficacy of priority molecules will be assessed against schistosomula, juvenile and adult worm lifecycle stages. These investigations will have long-term impacts by demonstrating how interdisciplinary methods can be used to accelerate the pace of anti-schistosomal drug discovery with actionable exemplar compounds (multi-stage activity, low micromolar potency) identified for entry into the hit-to-lead phase.

Role and Responsibility of Each Partner

BrightCore, Inc. (BC) will develop and train AI models from the co-assembled dataset of active and inactive compounds and run them on chemical libraries to find new compounds with potential anti-schistosomal activity. BC will develop and utilize domain specific, fine-tuned LLMs to generate compounds for confirmation with the trained ML models. ADMET (ADMET-AI based) testing will also be performed to increase quality of hits.

Aberystwyth University (AU) will conduct all ex vivo schistosome assays (e.g. schistosomula, juveniles and worms) on semi-automated phenotyping platforms that utilize high-content and multi-modal image analyses pipelines with parasite material derived from a lifecycle established in 2008.

University of Dundee (UoD) will be responsible for hit resynthesis or purchase as well as in vitro DMPK testing including metabolic stability, physicochemical characterization and cytotoxicity.  UoD will also be responsible for designing hit expansion plans for active series and synthesizing or purchasing analogs. UoD will be responsible for data management to ensure data preservation and integrity and compound management to ensure compounds are distributed to AU in a suitable format for ex vivo testing.