Novel Dual-Stage Antimalarials: Machine learning prediction, validation and evolution

项目摘要 具体来说，这项提案的重点是新的小分子，抑制血液和肝脏 疟疾感染的各个阶段。致病病原体-疟原虫属。- 2.41亿起案件 导致2020年62.7万人死亡。疟原虫属抗药性感染留下很少的好选择， 医生，并将感染者的生产力和生命置于危险之中。一个明确的理由已经提出了新的 通过发现和开发新的治疗策略来治疗这些感染。这些 最佳策略是双阶段，针对血液阶段进行治疗，针对肝脏阶段进行治疗。 预防。 该提案中的创新战略建立在机器学习模型技术的基础上， 预测新的双阶段抗疟小分子作为药物发现实体的巨大潜力。 这种计算方法为发现具有双功能的小分子疟疾寄生虫抑制剂奠定了基础， 我们仅在2022年报告了分期疗效。该方法首先从两个新的 抗疟小分子对疟原虫血液和肝脏阶段的体外功效 spp.感染和缺乏对培养的肝细胞的显著细胞毒性。这些分子来自于 使用高通量筛选数据训练的随机森林模型发现的命中集。的分子 是双阶段抗疟药的新型化学型的代表，因此， 调节在寄生虫整个生命周期中至关重要的新目标。这一初步的机器学习工作将 通过一系列模型类型和一个不同的更大的商业库来显著扩展， 一组新的热门化合物 两次经过验证的命中，符合体外功效和细胞毒性标准并在体外保持野生型 对一组耐药寄生虫菌株的疗效，将对关键分子特性进行分析，如小鼠 肝微粒体稳定性、水溶性和小鼠药代动力学特征。这些数据沿着 现有的体外疗效和细胞毒性评价将指导每一次命中的演变， 一种或多种具有复合特征的类似物，以使得能够在感染中进行下游体内功效评价 模型将利用药物化学和机器学习的新组合来提供这种 分子。

PROJECT SUMMARY Specifically, this proposal focuses on novel new small molecules that inhibit both the blood and liver stages of malaria infection. The causative pathogen – Plasmodium spp. – was responsible for 241,000,000 cases that resulted in 627,000 deaths in 2020. Plasmodium spp. drug-resistant infections leave few good choices for physicians and put at risk the productivity and the lives of those infected. A clear case has been made for new drugs to treat these infections through the discovery and development of novel therapeutic strategies. These strategies would optimally be dual stage, targeting the blood stage for treatment and the liver stage for prophylaxis. The innovative strategy in this proposal builds on the technology of machine learning models for the prediction of novel dual-stage antimalarial small molecules with significant potential as drug discovery entities. Such a computational approach to seed the discovery of small molecule malaria parasite inhibitors with dual- stage efficacy has only been reported by us in 2022. The approach begins with preliminary data around two novel antimalarial small molecules with demonstrated in vitro efficacy versus both blood and liver stages of Plasmodium spp. infection and a lack of significant cytotoxicity to cultured liver cells. These molecules were derived from a set of hits discovered with a random forest model trained with high-throughput screening data. The molecules are representative of novel chemotypes for dual-stage antimalarials and, thus, offer a high probability of modulating new targets that are critical throughout the parasite’s lifecycle. This initial machine learning effort will be significantly expanded with a range of model types and a different and larger commercial library to predict a set of new hit compounds. Two validated hits, meeting in vitro efficacy and cytotoxicity criteria and maintaining wild type in vitro efficacy versus a set of drug-resistant parasite strains, will be profiled for key molecular properties such as mouse liver microsomal stability, aqueous solubility, and mouse pharmacokinetic profile. These data along with the existing in vitro efficacy and cytotoxicity evaluations will guide the evolution of each hit with a goal of preparing one or more analogs with a composite profile to enable downstream in vivo efficacy evaluation in infection models. A novel combination of medicinal chemistry and machine learning will be leveraged to afford such molecules.