Advancing material property prediction: using physics-informed machine learning models for viscosity

In materials science, accurately computing properties like viscosity, melting point, and glass transition temperatures solely through physics-based models is challenging. Data-driven machine learning (ML) also poses challenges in constructing ML models, especially in the material science domain where data is limited. To address this, we integrate physics-informed descriptors from molecular dynamics (MD) simulations to enhance the accuracy and interpretability of ML models. Our current study focuses on accurately predicting viscosity in liquid systems using MD descriptors. In this work, we curated a comprehensive dataset of over 4000 small organic molecules' viscosities from scientific literature, publications, and online databases. This dataset enabled us to develop quantitative structure-property relationships (QSPR) consisting of descriptor-based and graph neural network models to predict temperature-dependent viscosities for a wide range of viscosities. The QSPR models reveal that including MD descriptors improves the prediction of experimental viscosities, particularly at the small data set scale of fewer than a thousand data points. Furthermore, feature importance tools reveal that intermolecular interactions captured by MD descriptors are most important for viscosity predictions. Finally, the QSPR models can accurately capture the inverse relationship between viscosity and temperature for six battery-relevant solvents, some of which were not included in the original data set. Our research highlights the effectiveness of incorporating MD descriptors into QSPR models, which leads to improved accuracy for properties that are difficult to predict when using physics-based models alone or when limited data is available.

A Computational Physics-based Approach to Predict Unbound Brain-to-Plasma Partition Coefficient, Kp,uu

The blood-brain barrier (BBB) plays a critical role in preventing harmful endogenous and exogenous substances from penetrating the brain. Optimal brain penetration of small-molecule central nervous system (CNS) drugs is characterized by a high unbound brain/plasma ratio (K_p,uu). While various medicinal chemistry strategies and in silico models have been reported to improve BBB penetration, they have limited application in predicting K_p,uu directly. We describe a physics-based computational approach, a quantum mechanics (QM)-based energy of solvation (E-sol), to predict K_p,uu. Prospective application of this method in internal CNS drug discovery programs highlights the utility and accuracy of this new method, which showed a categorical accuracy of 79% and an R² of 0.61 from a linear regression model.

Epik: pKa and Protonation State Prediction through Machine Learning

Epik version 7 is a software program that uses machine learning for predicting the pK_a values and protonation state distribution of complex, druglike molecules. Using an ensemble of atomic graph convolutional neural networks (GCNNs) trained on over 42,000 pK_a values across broad chemical space from both experimental and computed origins, the model predicts pK_a values with 0.42 and 0.72 pK_a unit median absolute and root mean square errors, respectively, across seven test sets. Epik version 7 also generates protonation states and recovers 95% of the most populated protonation states compared to previous versions. Requiring on average only 47 ms per ligand, Epik version 7 is rapid and accurate enough to evaluate protonation states for crucial molecules and prepare ultra-large libraries of compounds to explore vast regions of chemical space. The simplicity and time required for the training allow for the generation of highly accurate models customized to a program's specific chemistry.

Abstract 2570: Discovery of potent, selective, and orally available WEE1 inhibitors that demonstrate increased DNA damage and mitosis in tumor cells leading to tumor regression in vivo

WEE1 inhibits the activation of both CDK1 and CDK2 through phosphorylation of Tyr15, allowing DNA damage repair before entering mitosis, thereby regulating the cell cycle in S and G2/M phases. Inhibition of WEE1 could result in premature progression through the G2/M cell cycle checkpoint with unresolved DNA damage, leading to mitotic catastrophe and cell death. Small molecule WEE1 inhibitors, such as AZD1775 and Zn-C3, are currently being evaluated in the clinic and have demonstrated promising efficacy in solid tumors including ovarian, colon, and uterine carcinoma.

By applying Schrödinger’s computational platform including Free Energy Perturbation (FEP) and Protein FEP, we have identified novel, potent, and highly selective WEE1 inhibitors with IC50 values in the low nanomolar range in a biochemical kinase activity assay and cellular target engagement (CDK1 pTyr15) IC50s of 100 - 300 nM in A427 and OVCAR3 cell lines. The compounds also show potent anti-proliferative activity in over 20 breast and ovarian tumor cell lines, including cell lines insensitive to PARP inhibitors. The compounds demonstrate superior kinase selectivity compared to AZD1775 and Zn-C3 in a broad kinase panel with >450 kinases (ScanMAX). In addition, the compounds show desirable ADME properties and PK profiles in preclinical species. Based on in vitro CYP3A4 TDI assay performance (kinact/KI), we have reduced the potential for drug-drug interaction liabilities compared to AZD1775. In the A427 xenograft model, our WEE1 inhibitors demonstrate dose-dependent tumor growth inhibition and tumor regression at high doses. Anti-tumor activity is also demonstrated in additional tumor models, including OVCAR3 and HCC1806 xenograft models. The established PK-PD relationship shows sustained target engagement (pCDK1), increased DNA damage (gH2AX) and mitosis (pHH3). We demonstrate that hematological adverse effects can be mitigated by dosing holidays in xenograft tumor models while maintaining anti-tumor activity. Notably, our compound shows more sustained anti-tumor activity with dosing holidays compared to AZD1775, which we believe is attributable to the prolonged and higher exposure in tumor and plasma with our compound. In the A427 non small-cell lung cancer xenograft model, following 3 dosing holiday cycles at high doses, tumor eradication was maintained after treatment was stopped. In summary, we have identified novel, potent and exquisitely selective WEE1 kinase inhibitors that demonstrate robust anti-tumor activity and sustained target engagement in tumor models. The compound’s anti-tumor effects are maintained with dosing holidays while allowing full recovery of mechanism-based hematological effects.

Citation Format: Shaoxian Sun, Sarah Silvergleid, Aleksey I. Gerasyuto, Jiashi Wang, Robert D. Pelletier, Andrew Placzek, Jennifer L. Knight, Anthony Clark, Hamish Wright, Wu Yin, Jackson Chief Elk, Jeff Bell, Pieter H. Bos, Nicholas A. Boyles, Eric Therrien, Kristian Jensen, Karen Akinsanya. Discovery of potent, selective, and orally available WEE1 inhibitors that demonstrate increased DNA damage and mitosis in tumor cells leading to tumor regression in vivo [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 2570.

Abstract 1277: Discovery of novel CDC7 inhibitors that disrupt cell cycle dynamics and show anti-proliferative effects in cancer cells

Introduction: CDC7 is a serine/threonine protein kinase that phosphorylates the MCM2-7 helicase complex, a required step in DNA replication initiation. CDC7 has emerged as an attractive target for cancer treatment because of high expression in a number of tumors (e.g. ovarian, lung, and oral) which is thought to be linked to their proliferative capacity and ability to bypass DNA damage responses. Consistent with this, disruption of CDC7 activity in cancer cells results in delayed DNA replication, mitotic abnormalities and cell death whereas non-transformed, p53 wildtype cells are protected from cytotoxicity due to G1 cell cycle arrest. Due to the low ATP Km of CDC7, very potent inhibitor molecules are required to effectively block CDC7 activity and drive cancer cells into apoptosis. We have identified novel potent and selective CDC7 inhibitors targeting the ATP binding site that are active in biophysical, biochemical and cellular assays as well as in vivo CDX models.

Results: Our lead compounds show potent picomolar (pM) inhibition of CDC7 in a biochemical kinase activity assay, pM affinity in SPR assay and complete inhibition of MCM2 (S53) phosphorylation in COLO205, A427, MV-4-11 and SW48 cancer cell lines. In a broad kinase selectivity panel, the novel inhibitors showed good selectivity for CDC7 kinase. Mechanistic studies show that our CDC7 inhibitors induced apoptosis, disrupted DNA replication and cell cycle dynamics with accumulation of polyploid cells after 48 h of treatment of cancer cells with minimal effects on human fibroblast cell lines. Our compounds have shown potent anti-proliferative and cytotoxic effects in a panel of more than a 100 cancer cell lines of varying origin including COLO205, SW48, A427, MOLM-13, and SUM149. Comparison of CDC7 inhibitors with other oncology drugs in a panel of cancer cell lines revealed a unique mechanism of action. In vivo, our compounds reduced tumor cell MCM2 (S53) phosphorylation in the mouse COLO205 xenograft model and showed strong tumor growth inhibition. We have also examined the effect of CDC7 inhibitors on cancer cell proliferation in combination with other anti-cancer agents, including other DNA damage response (DDR) targeting agents.

Conclusions: We have identified novel potent ATP-competitive CDC7 inhibitors that show target engagement in cells and CDX tumors and have shown strong inhibition of cancer cell proliferation in vitro and tumor growth in vivo. CDC7 inhibitors show promise for use in combination with other targeted therapies for the treatment of cancers of varying origin.

Citation Format: Lyuben Tsvetkov, Adam Levinson, Xianhai Huang, Sayan Mondal, Jeff Bell, Lin Tang, Robert Pelletier, Karen Dingley, Nick Boyles, Jackson Chief Elk, Leah Frye, Alan Futran, Phani Ghanakota, Jeremy Greenwood, George Lai, Sarah Silvergleid, Wu Yin, Hamish Wright, Karen Akinsanya, Wayne Tang, Kristian Jensen. Discovery of novel CDC7 inhibitors that disrupt cell cycle dynamics and show anti-proliferative effects in cancer cells [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 1277.

β-Cyclodextrin at the Water/1-Bromobutane Interface: Molecular Insight into Reverse Phase Transfer Catalysis

Molecular insight into the role of β-cyclodextrin (βCD) as a phase transfer catalyst at the liquid/liquid interface is obtained by molecular dynamics simulations of the structure and dynamics of βCD adsorbed at the interface between water and 1-bromobutane. In particular, we consider the structure and dynamics of the water and bromobutane molecules inside the βCD cavity and compare them with the behavior when βCD is dissolved in bulk water. βCD is preferentially oriented at the interface, with the cavity opening along the interface normal. While in bulk water the cavity includes 6-8 water molecules that are relatively mobile with short residence time, at the interface the cavity is mostly dehydrated and includes a single bromobutane molecule. This inclusion complex is stable in bulk water. The implication of this behavior for reverse phase transfer catalysis is discussed.