Accurate and Reliable Prediction of Relative Ligand Binding Potency in Prospective Drug Discovery by Way of a Modern Free-Energy Calculation Protocol and Force Field

TLDR: An approach to designing tight-binding ligands with a substantial reduction in false positives relative to compounds synthesized on the basis of other computational or medicinal chemistry approaches is reported, demonstrating the robustness and broad range of applicability of this approach, which can be used to drive decisions in lead optimization.

Abstract: Designing tight-binding ligands is a primary objective of small-molecule drug discovery. Over the past few decades, free-energy calculations have benefited from improved force fields and sampling algorithms, as well as the advent of low-cost parallel computing. However, it has proven to be challenging to reliably achieve the level of accuracy that would be needed to guide lead optimization (∼5× in binding affinity) for a wide range of ligands and protein targets. Not surprisingly, widespread commercial application of free-energy simulations has been limited due to the lack of large-scale validation coupled with the technical challenges traditionally associated with running these types of calculations. Here, we report an approach that achieves an unprecedented level of accuracy across a broad range of target classes and ligands, with retrospective results encompassing 200 ligands and a wide variety of chemical perturbations, many of which involve significant changes in ligand chemical structures. In addition, we have applied the method in prospective drug discovery projects and found a significant improvement in the quality of the compounds synthesized that have been predicted to be potent. Compounds predicted to be potent by this approach have a substantial reduction in false positives relative to compounds synthesized on the basis of other computational or medicinal chemistry approaches. Furthermore, the results are consistent with those obtained from our retrospective studies, demonstrating the robustness and broad range of applicability of this approach, which can be used to drive decisions in lead optimization.

Figures

Table 3. Error distribution for all 330 perturbations. The distribution is approximately Gaussian with some fattening of the tail of the distribution beyond 2.5 kcal/mol. The fitting of the error distribution by a Gaussian function with the same RMSE is given in SI Fig. S1. The observed percentage is the percentage of FEP perturbations found to be accurate within the specified absolute error. The anticipated percentage is the percentage of FEP perturbations that would be expected given an underlying root-mean-square error of 1.1 kcal/mol and an ideal Gaussian error distribution.

Figure 4. Representative examples of different types of interactions captured by FEP (see text for details).

Figure 3. Correlation between the FEP predicted binding free energies and the experimental data for all 8 systems studied. FEP predicted binding free energies for most of the ligands are within 1.0 kcal/mol from their experimental values, and only 9 of 199 studied ligands deviate from their experimental free energies by more than 2 kcal/mol.

Figure 5: Histograms showing the distribution of experimental values for compounds in Project I that were predicted by FEP to have a pKi greater than 8 (dark gray), predicted by FEP to have a pKi less than 8 (medium gray), and those that were not computationally predicted prior to being assayed (light gray). The numbers above the bars correspond to the actual number of compounds that were assayed. Approximately 14% of the compounds that were chosen to be synthesized and assayed without guidance from FEP had a pKi >8, while 71% of the compounds predicted by FEP to have pKi >8 had an experimental pKi >8. Key to labels: TP=true positive, FN=false negative, FP=false positive, TN=true negative.

Figure 2: (a) The whole FEP workflow for protein-ligand binding affinities calculations. The key steps in the workflow, including the mutation graph generation, the REST region assignment, cycle closure convergence and error estimates and other key steps are explicitly shown in the figure. Note here that, the FEP/REST enhanced sampling method and the cycle closure error estimate were reported in prior publications,37,38 important related work about the perturbation graph generation was also reported in a prior publication,41 and the remaining components of the workflow are novel. (b) An example of a mapping of a perturbation space onto a set of pathways for thrombin ligands generated from the workflow. Each line represents two full FEP calculations, one

Table 1. Statistics of the errors for solvation free energy results reported by Shivakumar et al.32 using OPLS2.1, OPLS2005, AM1-BCC/GAFF, and ChelpG/CHARMM-MSI for the test set of 239 small molecules.

Full text

UC Irvine
UC Irvine Previously Published Works
Title
Accurate and reliable prediction of relative ligand binding potency in prospective drug
discovery by way of a modern free-energy calculation protocol and force field.
Permalink
https://escholarship.org/uc/item/5061663w
Journal
Journal of the American Chemical Society, 137(7)
ISSN
0002-7863
Authors
Wang, Lingle
Wu, Yujie
Deng, Yuqing
et al.
Publication Date
2015-02-01
DOI
10.1021/ja512751q
Peer reviewed
eScholarship.org Powered by the California Digital Library
University of California

1
Accurate and Reliable Prediction of Relative Ligand Binding Potency in Prospective Drug
Discovery by way of a Modern Free Energy Calculation Protocol and Force Field
Authors: Lingle Wang,
1
Yujie Wu,
1
Yuqing Deng,
1
Byungchan Kim,
1
Levi Pierce,
1
Goran Krilov,
1
Dmitry Lupyan,
1
Shaughnessy Robinson,
1
Markus K. Dahlgren,
1
Jeremy
Greenwood,
1
Donna L. Romero,
2
Craig Masse,
2
Jennifer L. Knight,
1
Thomas
Steinbrecher,
1
Thijs Beuming,
1
Wolfgang Damm,
1
Ed Harder,
1
Woody Sherman,
1
Mark
Brewer,
1
Ron Wester,
2
Mark Murcko,
1
Leah Frye,
1
Ramy Farid,
1
Teng Lin,
1
David L.
Mobley,
5
William L. Jorgensen,
4
Bruce J. Berne,
3
Richard A. Friesner,
3
Robert Abel
1,*
1
Schrödinger, Inc., 120 West 45th Street, New York, NY 10036, United States
2
Nimbus Discovery, 25 First Street, Suite 404 Cambridge, MA 02141, United States
3
Department of Chemistry, Columbia University, 3000 Broadway, New York, NY,
10027, United States
4
Department of Chemistry, Yale University, New Haven, CT 06520, United States
5
Departments of Pharmaceutical Sciences and Chemistry, University of California—
Irvine, Irvine, CA 92697, United States
*
Corresponding author email: robert.abel@schrodinger.com

2
Abstract
Designing tight binding ligands is a primary objective of small molecule drug discovery.
Over the past few decades, free energy calculations have benefited from improved force
fields and sampling algorithms, as well as the advent of low cost parallel computing.
However, it has proven to be challenging to reliably achieve the level of accuracy that
would be needed to guide lead optimization (~5X in binding affinity) for a wide range of
ligands and protein targets. Not surprisingly, widespread commercial application of free
energy simulations has been limited due to the lack of large-scale validation coupled with
the technical challenges traditionally associated with running these types of calculations.
Here, we report an approach that achieves an unprecedented level of accuracy across a
broad range of target classes and ligands, with retrospective results encompassing 200
ligands and a wide variety of chemical perturbations, many of which involve significant
changes in ligand chemical structures. In addition, we have applied the method in
prospective drug discovery projects and found a significant improvement in the quality of
the compounds synthesized that have been predicted to be potent. Compounds predicted
to be potent by this approach have a substantial reduction in false positives relative to
compounds synthesized based on other computational or medicinal chemistry
approaches. Furthermore, the results are consistent with those obtained from our
retrospective studies, demonstrating the robustness and broad range of applicability of
this approach, which can be used to drive decisions in lead optimization.

3
Introduction
Protein-ligand binding is central to both biological function and pharmaceutical activity.
Some ligands simply inhibit protein function, while others induce protein conformational
changes and hence can modulate key cell signaling pathways. In either case, achieving a
desired therapeutic effect is dependent upon the magnitude of the binding affinity of the
ligand to the target receptor. Designing tight binding ligands while maintaining the other
ligand properties required for safety and biological efficacy is a primary objective of
small molecule drug discovery projects.
A principal goal of computational chemistry and computer-aided drug design (CADD) is
therefore the accurate prediction of protein-ligand free energies of binding (i.e., binding
affinities).
1,2
The most rigorous approach to this problem is free energy simulation. A
variety of free energy simulation methods, such as free energy perturbation (FEP),
thermodynamic integration (TI), and lambda dynamics employ an analysis of atomistic
molecular dynamics or Monte Carlo simulations to determine the free energy difference
between two related ligands via either a chemical or alchemical path.
3-9
In drug discovery
lead optimization applications, the calculation of relative binding affinities (i.e., the
relative difference in binding energy between two compounds) is generally the quantity
of interest and affords significant reduction in computational effort as compared to
absolute binding free energy calculations.

4
Nearly three decades have passed since the initial applications of free energy methods to
the calculation of protein-ligand binding affinities were first reported by the Jorgensen,
McCammon, and Kollman groups.
10-15
Subsequent efforts since that original seminal
work have reported anecdotal results for a small number of protein-ligand complexes, but
suffered from a lack of computing power and inadequacies in both sampling algorithms
and molecular mechanics force fields.
1,5,8
As a result, use of free energy calculations was
limited in an industrial drug discovery setting, where high throughput, predictive
accuracy, and robustness are required to make a significant impact.
In recent years, FEP calculations have benefitted from improved force fields, new
sampling algorithms, and the emergence of low cost parallel computing, which have
resulted in the level of accuracy and turnaround time needed to impact lead optimization
efforts, as demonstrated in several academic projects.
1,5,8,16-20
However, it has not been
demonstrated that highly accurate results can be achieved reliably across a wide range of
ligands and protein targets, as would be needed for the method to be useful in industrial
pharmaceutical research programs.
Here, we report an FEP protocol that enables highly accurate affinity predictions across a
broad range of ligands and target classes (over 200 ligands and 10 targets). The ligand
perturbations include a wide-range of chemical modifications that are typically seen in
medicinal chemistry efforts, with modifications of up to ten heavy atoms routinely
included. Critically, we have applied the method in 8 prospective discovery projects to
date, with the results from two of those projects disclosed in this work. The high level of

Frequently asked questions

Q1. What is the effect of the amino group in the second ligand?

When mutating the amino-group into a methoxy-group in second ligand (right panel), the experimental binding free energy is observed to decrease by 0.5 kcal/mol, presumably due to the loss of hydrogen bonding interactions. 

Q2. What is the primary objective of small molecule drug discovery?

Designing tight binding ligands while maintaining the other ligand properties required for safety and biological efficacy is a primary objective of small molecule drug discovery projects. 

Q3. What are the contributions in "Accurate and reliable prediction of relative ligand binding potency in prospective drug discovery by way of a modern free-energy calculation protocol and force field" ?

Free energy perturbation ( FEP ) has been used for protein-ligand binding this paper. 

Q4. How do the authors obtain torsional parameters for a druglike molecule?

The torsional parameters are obtained by constructing model compounds containing the relevant torsional structures,and fitting the parameters to quantum chemical data computed at the LMP2/cc-pVTZ(-f) level of theory, which has been shown to yield accurate relative conformational energies for the systems being modeled. 

Q5. What is the experimental error for the binding free energies of two compounds?

Assuming the experimental measurements for the binding free energies of two compounds are independent, ie, , then the experimental error for therelative binding free energies between two compounds is about 0.4-0.7 kcal/mol. 

Q6. How many perturbations per day can be completed with the protocol described in this work?

For a “typical” FEP calculation (~6,000 atoms in the protein) with the protocol described in this work, 4 perturbations per day can be completed using 8 commodity Nvidia GTX-780 GPUs, making it feasible to evaluate thousands of molecules per year in the context of a drug discovery program with compute resources that are well within the reach of both academic institutions and commercial enterprises. 

Q7. How many compounds were predicted to have a pKi 8?

37 compounds with pKi predictions of ≤8 were not synthesized; the true negative rate in Project II was 75% based on results for 4 compounds predicted to have a pKi ≤8 and were subsequently synthesized. 

Q8. What is the effect of adding a chloro-group to the first ligand?

With the addition of a chloro-group at the meta position of the phenyl ring in the first ligand (left panel), the high energy water molecule in the S1 pocket isdisplaced, resulting in a more favorable binding free energy for the second ligand. 

Q9. How many purchasable drug-like compounds are required to represent the diversity of even this?

Their analysis of one million purchasable drug-like compounds indicates that on the order of tens of thousands of such compounds are required to represent the diversity of even this limited chemical space. 

Q10. What is the common modification of ligands?

The ligand perturbations include a wide-range of chemical modifications that are typically seen in medicinal chemistry efforts, with modifications of up to ten heavy atoms routinely included. 

Q11. What is the effect of the loss of the mobility of the amino group on binding to the receptor?

As such, the authors believe the loss of the mobility of this group upon binding to the receptor may lead to entropic penalties weakening the binding of the ligand to the receptor. 

Q12. How many compounds were scored with FEP?

At this stage of the project and over a period of several months, 195 compounds were prospectively scored with FEP, and 22 were synthesized and assayed. 

Q13. What is the value of the FEP scoring in practical applications?

This final case illustrates another point regarding the value of FEP scoring in practical applications: the maximum size of the perturbations that can be reliably treated is of equal significance to obtaining predictive correlation and small RMS errors. 

Q14. What is the value of the FEP scoring in practical applications?

This final case illustrates another point regarding the value of FEP scoring in practical applications: the maximum size of the perturbations that can be reliably treated is of equal significance to obtaining predictive correlation and small RMS errors. 

Q15. What is the effect of FEP scoring on the synthesis of tight binding molecules?

the observed 6-fold enrichment in the synthesis of tight binding molecules provides suggestive evidence FEP scoring provides a substantial reduction in false positives relative to compounds synthesized based on other approaches. 

Q16. What is the effect of FEP scoring on the synthesis of tight binding molecules?

the observed 6-fold enrichment in the synthesis of tight binding molecules provides suggestive evidence FEP scoring provides a substantial reduction in false positives relative to compounds synthesized based on other approaches. 

Q17. What is the goal of computational chemistry and computer-aided drug design?

A principal goal of computational chemistry and computer-aided drug design (CADD) is therefore the accurate prediction of protein-ligand free energies of binding (i.e., binding affinities). 

Q18. What is the average score for the series reported in table 2?

The FEP scoring weighted average R-value obtained for the series reported in table 2 is 0.75, for MM-GB/SA it is 0.35, and for Glide SP it is 0.29. 

Q19. How long would it take to prepare the calculations?

The 16 separate calculations shown in Fig. 2b can be prepared in approximately 30 minutes, whereas manual setup without a graphical user interface and automated mapping protocols would take significantly longer. 

Q20. How long would it take to prepare the calculations?

The 16 separate calculations shown in Fig. 2b can be prepared in approximately 30 minutes, whereas manual setup without a graphical user interface and automated mapping protocols would take significantly longer. 

Q21. How many bond corrections have been developed for challenging chemistries?

28,29 Ligand atomic partial charges are computed via a CM1A-BCC methodology30,31 where a substantial number of bond charge corrections for challenging chemistries have been developed.