Computational Investigation of Drug Phototoxicity: Photosafety Assessment, Photo-Toxophore Identification, and Machine Learning

Abstract

One of the most appreciated capabilities of computational toxicology is to support the design of pharmaceuticals with reduced toxicological hazard. To this end, we have strengthened our drug photosafety assessments by applying novel computer models for the anticipation of in vitro phototoxicity and human photosensitization. These models are typically used in pharmaceutical discovery projects as part of the compound toxicity assessments and compound optimization methods. To ensure good data quality and aiming at models with global applicability we separately compiled and curated highly chemically diverse data sets from 3T3 NRU phototoxicity reports (450 compounds) and clinical photosensitization alerts (1419 compounds) which are provided as supplements. The latter data gives rise to a comprehensive list of explanatory fragments for visual guidance, termed phototoxophores, by application of a Bayesian statistics approach. To extend beyond the domain of well sampled fragments we applied machine learning techniques based on explanatory descriptors such as pharmacophoric fingerprints or, more important, accurate electronic energy descriptors. Electronic descriptors were extracted from quantum chemical computations at the density functional theory (DFT) level. Accurate UV/vis spectral absorption descriptors and pharmacophoric fingerprints turned out to be necessary for predictive computer models, which were both derived from Deep Neural Networks but also the simpler Random Decision Forests approach. Model accuracies of 83–85% could typically be reached for diverse test data sets and other company in-house data, while model sensitivity (the capability of correctly detecting toxicants) was even better, reaching 86%–90%. Importantly, a computer model-triggered response-map allowed for graphical/chemical interpretability also in the case of previously unknown phototoxophores. The photosafety models described here are currently applied in a prospective manner for the hazard identification, prioritization, and optimization of newly designed molecules.

Toxicity prediction based on artificial intelligence: A multidisciplinary overview

“…The exposure of large groups of people encourages the convenience of implementing methodologies capable of predicting phototoxicity. In the early stages of development, Schmidt et al 104 developed novel computer models using random decision forests and DNNs capable of anticipating in vivo phototoxicity and human photosensitization. Data sets from 3T3 neutral red uptake phototoxicity reports (450 compounds) and clinical photosensitization alerts (1419 compounds) were used for training, attaining model accuracies of 83% to 85% and sensitivities (capability of correctly detecting toxicants) between 86% and 90%.…”

Application of Deep Neural Network Models in Drug Discovery Programs

“…[11,14,[22][23] Successful examples include inhibition of cytochrome P450s, [24] metabolic lability, [11] and toxicity predictions. [25][26][27][28] However, whether joint training with multiple endpoints improves predictivity, depends on the correlation of activities for model building and structural overlap. [14][15][16] Outline.…”

Machine learning prediction of UV–Vis spectra features of organic compounds related to photoreactive potential

“…The UV–Vis absorption spectrum is a key physical property of an organic compound that determines many of its optoelectronic properties and photochemical reactivity. In the human body, the combined effect of an external chemical compound (e.g., plant toxins, phytomedicines, cosmetics, agrochemicals, food additives, dyes, drugs) and exposure to light, especially ultraviolet and visible radiation may give rise to an acute unwanted response in the skin or retina, which is called chemical phototoxicity 1 , 2 .…”

“…Phototoxicity is strongly related to molecular photochemistry and photostability 2 . The optimization of ADME-Tox parameters (absorption, distribution, metabolism, excretion, and toxicology) using high-throughput tools is of great importance in drug discovery 12 , and ML approaches can be used to rationalize and predict phototoxicity, representing a valuable strategy for reducing experimental tests, if an acceptable level of accuracy of the developed models is ensured.…”