Predicting Volume of Distribution in Humans: Performance of In Silico Methods for a Large Set of Structurally Diverse Clinical Compounds.

Volume of distribution at steady state (VD,ss) is one of the key pharmacokinetic parameters estimated during the drug discovery process. Despite considerable efforts to predict VD,ss, accuracy and choice of prediction methods remain a challenge, with evaluations constrained to a small set (<150) of compounds. To address these issues, a series of in silico methods for predicting human VD,ss directly from structure were evaluated using a large set of clinical compounds. Machine learning (ML) models were built to predict VD,ss directly and to predict input parameters required for mechanistic and empirical VD,ss predictions. In addition, log D, fraction unbound in plasma (fup), and blood-to-plasma partition ratio (BPR) were measured on 254 compounds to estimate the impact of measured data on predictive performance of mechanistic models. Furthermore, the impact of novel methodologies such as measuring partition (Kp) in adipocytes and myocytes (n = 189) on VD,ss predictions was also investigated. In predicting VD,ss directly from chemical structures, both mechanistic and empirical scaling using a combination of predicted rat and dog VD,ss demonstrated comparable performance (62%-71% within 3-fold). The direct ML model outperformed other in silico methods (75% within 3-fold, r 2 = 0.5, AAFE = 2.2) when built from a larger data set. Scaling to human from predicted VD,ss of either rat or dog yielded poor results (<47% within 3-fold). Measured fup and BPR improved performance of mechanistic VD,ss predictions significantly (81% within 3-fold, r 2 = 0.6, AAFE = 2.0). Adipocyte intracellular Kp showed good correlation to the VD,ss but was limited in estimating the compounds with low VD,ss SIGNIFICANCE STATEMENT: This work advances the in silico prediction of VD,ss directly from structure and with the aid of in vitro data. Rigorous and comprehensive evaluation of various methods using a large set of clinical compounds (n = 956) is presented. The scale of techniques evaluated is far beyond any previously presented. The novel data set (n = 254) generated using a single protocol for each in vitro assay reported in this study could further aid in advancing VD,ss prediction methodologies.

Accelerating Therapeutics for Opportunities in Medicine: A Paradigm Shift in Drug Discovery.

Conventional drug discovery is long and costly, and suffers from high attrition rates, often leaving patients with limited or expensive treatment options. Recognizing the overwhelming need to accelerate this process and increase success, the ATOM consortium was formed by government, industry, and academic partners in October 2017. ATOM applies a team science and open-source approach to foster a paradigm shift in drug discovery. ATOM is developing and validating a precompetitive, preclinical, small molecule drug discovery platform that simultaneously optimizes pharmacokinetics, toxicity, protein-ligand interactions, systems-level models, molecular design, and novel compound generation. To achieve this, the ATOM Modeling Pipeline (AMPL) has been developed to enable advanced and emerging machine learning (ML) approaches to build models from diverse historical drug discovery data. This modular pipeline has been designed to couple with a generative algorithm that optimizes multiple parameters necessary for drug discovery. ATOM's approach is to consider the full pharmacology and therapeutic window of the drug concurrently, through computationally-driven design, thereby reducing the number of molecules that are selected for experimental validation. Here, we discuss the role of collaborative efforts such as consortia and public-private partnerships in accelerating cross disciplinary innovation and the development of open-source tools for drug discovery.

AMPL: A Data-Driven Modeling Pipeline for Drug Discovery.

One of the key requirements for incorporating machine learning (ML) into the drug discovery process is complete traceability and reproducibility of the model building and evaluation process. With this in mind, we have developed an end-to-end modular and extensible software pipeline for building and sharing ML models that predict key pharma-relevant parameters. The ATOM Modeling PipeLine, or AMPL, extends the functionality of the open source library DeepChem and supports an array of ML and molecular featurization tools. We have benchmarked AMPL on a large collection of pharmaceutical data sets covering a wide range of parameters. Our key findings indicate that traditional molecular fingerprints underperform other feature representation methods. We also find that data set size correlates directly with prediction performance, which points to the need to expand public data sets. Uncertainty quantification can help predict model error, but correlation with error varies considerably between data sets and model types. Our findings point to the need for an extensible pipeline that can be shared to make model building more widely accessible and reproducible. This software is open source and available at: https://github.com/ATOMconsortium/AMPL.

Simulation of lipid bilayer self-assembly using all-atom lipid force fields.

In this manuscript we expand significantly on our earlier communication by investigating the bilayer self-assembly of eight different types of phospholipids in unbiased molecular dynamics (MD) simulations using three widely used all-atom lipid force fields. Irrespective of the underlying force field, the lipids are shown to spontaneously form stable lamellar bilayer structures within 1 microsecond, the majority of which display properties in satisfactory agreement with the experimental data. The lipids self-assemble via the same general mechanism, though at formation rates that differ both between lipid types, force fields and even repeats on the same lipid/force field combination. In addition to zwitterionic phosphatidylcholine (PC) and phosphatidylethanolamine (PE) lipids, anionic phosphatidylserine (PS) and phosphatidylglycerol (PG) lipids are represented. To our knowledge this is the first time bilayer self-assembly of phospholipids with negatively charged head groups is demonstrated in all-atom MD simulations.

A Parameterization of Cholesterol for Mixed Lipid Bilayer Simulation within the Amber Lipid14 Force Field.

The Amber Lipid14 force field is expanded to include cholesterol parameters for all-atom cholesterol and lipid bilayer molecular dynamics simulations. The General Amber and Lipid14 force fields are used as a basis for assigning atom types and basic parameters. A new RESP charge derivation for cholesterol is presented, and tail parameters are adapted from Lipid14 alkane tails. 1,2-Dimyristoyl-sn-glycero-3-phosphocholine (DMPC), 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), and 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) bilayers are simulated at a range of cholesterol contents. Experimental bilayer structural properties are compared with bilayer simulations and are found to be in good agreement. With this parameterization, another component of complex membranes is available for molecular dynamics with the Amber Lipid14 force field.

All-atom lipid bilayer self-assembly with the AMBER and CHARMM lipid force fields.

This communication reports the first example of spontaneous lipid bilayer formation in unbiased all-atom molecular dynamics (MD) simulations. Using two different lipid force fields we show simulations started from random mixtures of lipids and water in which four different types of phospholipids self-assemble into organized bilayers in under 1 microsecond.

Lipid14: The Amber Lipid Force Field.

The AMBER lipid force field has been updated to create Lipid14, allowing tensionless simulation of a number of lipid types with the AMBER MD package. The modular nature of this force field allows numerous combinations of head and tail groups to create different lipid types, enabling the easy insertion of new lipid species. The Lennard-Jones and torsion parameters of both the head and tail groups have been revised and updated partial charges calculated. The force field has been validated by simulating bilayers of six different lipid types for a total of 0.5 µs each without applying a surface tension; with favorable comparison to experiment for properties such as area per lipid, volume per lipid, bilayer thickness, NMR order parameters, scattering data, and lipid lateral diffusion. As the derivation of this force field is consistent with the AMBER development philosophy, Lipid14 is compatible with the AMBER protein, nucleic acid, carbohydrate, and small molecule force fields.

LIPID11: a modular framework for lipid simulations using amber.

Accurate simulation of complex lipid bilayers has long been a goal in condensed phase molecular dynamics (MD). Structure and function of membrane-bound proteins are highly dependent on the lipid bilayer environment and are challenging to study through experimental methods. Within Amber, there has been limited focus on lipid simulations, although some success has been seen with the use of the General Amber Force Field (GAFF). However, to date there are no dedicated Amber lipid force fields. In this paper we describe a new charge derivation strategy for lipids consistent with the Amber RESP approach and a new atom and residue naming and type convention. In the first instance, we have combined this approach with GAFF parameters. The result is LIPID11, a flexible, modular framework for the simulation of lipids that is fully compatible with the existing Amber force fields. The charge derivation procedure, capping strategy, and nomenclature for LIPID11, along with preliminary simulation results and a discussion of the planned long-term parameter development are presented here. Our findings suggest that LIPID11 is a modular framework feasible for phospholipids and a flexible starting point for the development of a comprehensive, Amber-compatible lipid force field.