Eric Martin, PhD
(Co-Mentor: Ezra Tai, PhD)
Global Discovery Chemistry
Emeryville, California, United States
Rational Bioavailability Design using Physiologically-Based Pharmacokinetics Simulations
The Computer-Aided Drug Design group works at the interface of several scientific disciplines: chemistry, enzyme kinetics, pharmacokinetics, mathematical simulations, and statistics. Structure-based drug design (SBDD) is the main computational tool for improving binding affinity during lead optimization (LO), but does not address whole animal efficacy. We have recently developed methodology to perform reliable physiologically-based pharmacokinetic (PBPK) simulations of absorption, distribution, metabolism and elimination (ADME) not just for a few, well studied, late-stage compounds, but for entire medchem series, using only inputs computed from chemical structure. These whole-animal simulations only require about 15 rat PK studies for training, typically available early enough to impact LO. We analyze thousands of these simulations with partial least squares regression (PLS) based global sensitivity analysis to identify which few of perhaps a dozen compound properties dominate improved bioavailability across each entire series, as well as for each specific compound. Furthermore, the PLS function provides a very good approximation of the full PBPK simulation for that series.
While this work has extended PBPK modeling to LO, it does not yet suggest specific compounds to synthesize, and does not account for how whole-animal ADME interacts with molecular binding kinetics to determine target occupancy over time, and therefore efficacy. We now aim to integrate interactive SBDD and in silico compound evolution with multi-objective scoring functions that include both affinity and the PLS approximation of exposure, and to incorporate binding kinetics into the full PBPK simulations. This multi-scale systems biology approach to drug design would complete a path from SBDD, through binding-kinetics enabled PBPK and pharmacodynamics, enabling SBDD not just of ligand binding, but of in vivo efficacy. The goal is ambitious, yet the risks are minimized because the components are established methods, and we can draw on internal Novartis experts in all the required disciplines.
Improving bioavailability during lead optimization [webinar]
Martin, EJ, Daga P
Surrogate AutoShim: Predocking into a universal ensemble kinase receptor for three-dimensional activity prediction, very quickly, without a crystal structure
Martin, EJ, Sullivan DC,
J. Chem. Inf. Model, 2008; 48(4); 873-881.
Profile-QSAR: a novel meta-QSAR method that combines activities across the kinase family to accurately predict affinity, selectivity, and cellular activity.
Martin E, Mukherjee P, Sullivan D, Jansen J.
J Chem Inf Model. 2011 Aug 22;51(8):1942-56.
Click here for additional publications.