There have been some comparative studies done by independent researchers and evaluators wherein they have compared the Molsoft’s ICM with other molecular modeling tools in the market. We have summarized the findings and observations from such research articles below for your reference. The full text of the article can be accessed by clicking on the title of the paper or on the doi.
On Evaluating Molecular-Docking Methods for Pose Prediction and Enrichment Factors
– Hongming Chen, Paul D. Lyne, Fabrizio Giordanetto, Timothy Lovell, and Jin Li
J. Chem. Inf. Model. 2006, 46, 1, 401–415 | https://doi.org/10.1021/ci0503255
Hongming Chen et al. reported a comparative analysis of the four most well-known, commercially available docking programs, FlexX, GOLD, GLIDE, and ICM, in terms of their ligand-docking and virtual-screening capabilities. In this study they have employed a dataset of 164 high-resolution protein-ligand complexes to determine the relative performance of each docking program in reproducing the native ligand conformation starting from SMILES strings, subsequently, the predicted binding modes obtained using each of the programs were used to characterize the docking accuracy by calculating the several parameters such as the average RMSD for all the top solutions from each docking program, RMSD of each docking solution and success rate, which is the percentage of successful solutions coming from the top solutions at the two different RMSD thresholds (<2.0 Å and <1.0 Å).
The scatter plot of the RMSD against the number of compounds docked using four docking programs is shown on the left-hand side. From the plot analysis, it is revealed that ICM has the lowest average RMSD matching with the native ligands i.e., 1.08 Å in comparison to the other three programs.
At the RMSD cutoff of 2.0 Å depicted by the green box in each plot, ICM performed well and identified 149/164 redock poses for compounds correctly with success rates of 91%.
At the more stringent RMSD cutoff of 1.0 Å, denoted by the magenta box in each plot, ICM again performed well, correctly docking 93/164 leading to success rates of 57%.
The summary of generated docking results was provided in the table as follows:
The latest versions of these four docking programs were also utilized to undertake virtual screening for 12 protein targets of therapeutic relevance, utilizing both publicly available structures and AstraZeneca in-house structures. Subsequently, the comparisons are made between the four programs’ abilities to accurately rank-order target-specific active compounds relative to alternative binders and nonbinders (decoys + randomly chosen compounds) and it was found at a sub-setting level of 10%, ICM produced the best results overall with an enrichment factor of 6.1 when averaged over all 12 targets as shown in the table:
Comparative Evaluation of Covalent Docking Tools
– Andrea Scarpino, György G. Ferenczy, and György M. Keserű
J. Chem. Inf. Model. 2018, 58, 7, 1441–1458 | https://doi.org/10.1021/acs.jcim.8b00228
Andrea Scarpino et al. have reported the comparative studies on the evaluation of covalent docking tools, in this study, they have compared the performance of six covalent docking tools, AutoDock4, CovDock, FITTED, GOLD, ICM-Pro, and MOE, for reproducing experimental binding modes in an unprecedently large and diverse set of covalent complexes. In this study they have reported that the experimental ligand pose could be reproduced by MolSoft’s ICM-Pro software at a two Angstrom RMSD cutoff in 62% of the cases and in the Top 10 conformations in 88% of the cases. In every test, including reproducing, the experimental binding modes, and scoring, MolSoft’s ICM-Pro software surpassed the competition. ICM also operated admirably with more difficult flexible ligands that contained 35 heavy atoms or more. They have also discovered that the non-covalent conventional docking method of ICM-Pro also provides an incredibly high rate of accurate predictions for covalent ligands to modified receptors (CYS/ALA mutations—86% near native).
The success rate of re-docking is depicted in the figures below (from Scarpino et al) for the top 1 pose and the top 10 poses, respectively.
Comparative Evaluation of 3D Virtual Ligand Screening Methods: Impact of the Molecular Alignment on Enrichment
– David Giganti, Hélène Guillemain, Jean-Louis Spadoni, Michael Nilges, Jean-François Zagury, and Matthieu Montes
J. Chem. Inf. Model. 2010, 50, 6, 992–1004 | https://doi.org/10.1021/ci900507g
David Giganti et al. have performed and reported the comparative analysis of 3D-Ligand-based and docking-based virtual screening employing the four different software i.e., Surflex-dock/Surflex-sim, FlexX/FlexS, ICM, and OMEGA-FRED/OMEGA-ROCS. The dataset of 11 targets (ADA, CDK2, DHFR, ER, FXA, HIVRT, NA, P38, THR, TK, TRP) was carefully selected from DUD dataset, which is a benchmarking data set designed for docking method evaluation containing annotated active compounds (from 30 to 120) for 40 targets, including 36 decoys for each active molecule issued from the ZINC database. They have compared the performance of all four programs in terms of molecular alignment accuracy, enrichment in active compounds, and enrichment in different chemotypes (scaffold hopping). It was found that, for both docking-BVLS and 3D-LBVLS, ICM displays the best performance in terms of molecular alignment and docking accuracy. This may be caused by a more effective treatment of the flexibility of the compounds in ICM via its biased probability Monte Carlo procedure. Similarly, ICM had the best global performance in enrichment.