Accutar Biotechnology Inc., 760 Parkside Ave., Brooklyn, NY 11226, USA.
Amgen Inc., 1 Amgen Center Dr., Thousand Oaks, CA 91320, USA.
Int J Mol Sci. 2019 Jul 10;20(14):3389. doi: 10.3390/ijms20143389.
Absorption, distribution, metabolism, and excretion (ADME) studies are critical for drug discovery. Conventionally, these tasks, together with other chemical property predictions, rely on domain-specific feature descriptors, or fingerprints. Following the recent success of neural networks, we developed Chemi-Net, a completely data-driven, domain knowledge-free, deep learning method for ADME property prediction. To compare the relative performance of Chemi-Net with Cubist, one of the popular machine learning programs used by Amgen, a large-scale ADME property prediction study was performed on-site at Amgen. For all 13 data sets, Chemi-Net resulted in higher R values compared with the Cubist benchmark. The median R increase rate over Cubist was 26.7%. We expect that the significantly increased accuracy of ADME prediction seen with Chemi-Net over Cubist will greatly accelerate drug discovery.
吸收、分布、代谢和排泄 (ADME) 研究对于药物发现至关重要。传统上，这些任务以及其他化学性质预测都依赖于特定于领域的特征描述符或指纹。在神经网络最近取得成功之后，我们开发了 Chemi-Net，这是一种完全由数据驱动、无领域知识的深度学习方法，用于 ADME 性质预测。为了比较 Chemi-Net 与 Amgen 使用的流行机器学习程序之一 Cubist 的相对性能，在 Amgen 现场进行了大规模的 ADME 性质预测研究。对于所有 13 个数据集，Chemi-Net 的 R 值都高于 Cubist 基准。Chemi-Net 相对于 Cubist 的中位数 R 值增长率为 26.7%。我们预计，Chemi-Net 在 ADME 预测方面的准确性显著提高将大大加快药物发现的速度。
