Ortega-Tenezaca Bernabe, Quevedo-Tumailli Viviana, Bediaga Harbil, Collados Jon, Arrasate Sonia, Madariaga Gotzon, Munteanu Cristian R, Cordeiro M Natália D S, González-Díaz Humbert
RNASA-IMEDIR, Computer Science Faculty, University of A Coruna, 15071 A Coruña, Spain
Universidad Estatal Amazónica UEA, Puyo, Pastaza, Ecuador
Curr Top Med Chem. 2020;20(25):2326-2337. doi: 10.2174/1568026620666200916122616.
By combining Machine Learning (ML) methods with Perturbation Theory (PT), it is possible to develop predictive models for a variety of response targets. Such combination often known as Perturbation Theory Machine Learning (PTML) modeling comprises a set of techniques that can handle various physical, and chemical properties of different organisms, complex biological or material systems under multiple input conditions. In so doing, these techniques effectively integrate a manifold of diverse chemical and biological data into a single computational framework that can then be applied for screening lead chemicals as well as to find clues for improving the targeted response(s). PTML models have thus been extremely helpful in drug or material design efforts and found to be predictive and applicable across a broad space of systems. After a brief outline of the applied methodology, this work reviews the different uses of PTML in Medicinal Chemistry, as well as in other applications. Finally, we cover the development of software available nowadays for setting up PTML models from large datasets.
通过将机器学习（ML）方法与微扰理论（PT）相结合，可以开发针对各种响应目标的预测模型。这种通常被称为微扰理论机器学习（PTML）建模的结合包括一组能够处理不同生物体、复杂生物或材料系统在多种输入条件下的各种物理和化学性质的技术。通过这样做，这些技术有效地将大量不同的化学和生物学数据整合到一个单一的计算框架中，然后该框架可用于筛选先导化合物以及寻找改善目标响应的线索。因此，PTML模型在药物或材料设计工作中非常有帮助，并且发现在广泛的系统空间中具有预测性和适用性。在简要概述所应用的方法之后，本文回顾了PTML在药物化学以及其他应用中的不同用途。最后，我们介绍了目前可用于从大型数据集中建立PTML模型的软件的开发情况。
