Novel deep learning strategy to better predict pharmacological properties of candidate drugs and focus discovery efforts

项目总结 协作药物发现公司(CDD)建议继续开发一种新的方法，基于 深度学习神经网络将分子编码成化学丰富的载体。在第1阶段，我们演示了 这种表示使计算模型能够更准确地预测 分子比最先进的模型更容易制造，因为它们不需要任何专家 决策或优化以实现高性能。在第二阶段，我们将利用这种前所未有的简单性 开发一个直观的软件包，这将首次使任何在 药物发现创建和运行自己的预测模型--而不依赖于专门的化学信息学 专业技术--但仍能达到或超过目前最好的技术的精确度。科学家致力于 在药物发现研究中，从学术实验室到大型制药公司都依赖于 预测相关药理性质的计算QSAR模型，并消除执行 对每个感兴趣的分子进行昂贵、耗时的分析(其中许多需要进行动物研究)。 改进的模型将使研究人员能够更有效地选择领先的候选系列，探索化学 周围空间导致更高效地生成新的IP，降低化合物推进的故障率 通过药物发现流水线，加快整个药物发现进程。这些好处将是 在大多数治疗领域得到广泛应用。 我们还计划将这项技术更进一步，利用我们丰富的化学矢量表示法来 使该软件能够创造性地建议新化合物(其没有出现在训练库中， 筛选文库或先导系列)同时在生物活性上超过先导候选者， ADME/Tox和PK检测。解决这个逆问题是计算医学的圣杯 化学，并具有革命性的药物发现的潜力。 好了！

PROJECT SUMMARY Collaborative Drug Discovery, Inc. (CDD) proposes to continue development of a novel approach based on deep learning neural networks to encode molecules into chemically rich vectors. In Phase 1 we demonstrated that this representation enables computational models that more accurately predict the chemical properties of molecules than state-of-the-art models, yet are also far simpler to build because they do not require any expert decisions or optimization to achieve high performance. In Phase 2 we will exploit this unprecedented simplicity to develop an intuitive software package that will for the first time enable any chemist or biologist working in drug discovery to create and run their own predictive models – without relying on specialized cheminformatics expertise – yet still achieve or exceed the accuracy of the best currently available techniques. Scientists engaged in drug discovery research from academic laboratories to large pharmaceutical companies rely on computational QSAR models to predict pharmacologically relevant properties and obviate the need to perform expensive, time-consuming assays (many of which require animal studies) for every molecule of interest. Improved models will enable researchers to select lead candidate series more effectively, explore chemical space around leads to generate novel IP more efficiently, reduce failure rates for compounds advancing through the drug discovery pipeline, and accelerate the entire drug discovery process. These benefits will be realized broadly across most therapeutic areas. We also plan to take the technology one step further, leveraging our chemically rich vector representation to enable the software to creatively suggest novel compounds (which do not appear in the training libraries, screening libraries, or lead series) that outperform the lead candidates simultaneously on bioactivity, ADME/Tox and PK assays . Solving this inverse problem is the Holy Grail of computational medicinal chemistry and has the potential to revolutionize drug discovery. !