The SAMPL9 host–guest blind challenge: an overview of binding free energy predictive accuracy

Abstract

We report the results of the SAMPL9 host-guest blind challenge for predicting binding free energies. The challenge focused on macrocycles from pillar[n]-arene and cyclodextrin host families, including WP6, and bCD...

Methodological and force field effects in the molecular dynamics-based prediction of binding free energies of host–guest systems

“…The two approaches yield a significant difference in the nominal ionic strength of the MD box. Experimental measurements in SAMPL host–guest challenges are done in general in PBS buffer at pH = 7.4 with an ionic strength of the order of 0.2 M, 7,40 neglecting the contributions of the host salt whose concentration is at least 3 orders of magnitude smaller than that of the buffer.…”

“…The SAMPL challenge (statistical assessment of the modeling of proteins and ligands) is a long-standing international initiative, started in 2008, aimed at innovating and advancing reliable predictive tools for drug design. 1–7 The challenges over the last decade have been mostly focused on host–guest systems as “widely accepted model systems to validate computational binding affinity methods and to gain insight into the physical chemistry of molecular recognition.” 2 Typical SAMPL hosts, such as cyclodextrins, octa-acid cavitands, 8 or pillar[ n ]arenes, 9,10 are characterized by a concave binding surface able to accommodate drug-size molecules with a wide range of affinities in aqueous solution. The latest SAMPL9 challenge overview 7 reported on the results of the challenge for 8 cyclodextrin–phenothiazine systems and 13 ligands of the WP6 pillar[6]arene host.…”

“…1–7 The challenges over the last decade have been mostly focused on host–guest systems as “widely accepted model systems to validate computational binding affinity methods and to gain insight into the physical chemistry of molecular recognition.” 2 Typical SAMPL hosts, such as cyclodextrins, octa-acid cavitands, 8 or pillar[ n ]arenes, 9,10 are characterized by a concave binding surface able to accommodate drug-size molecules with a wide range of affinities in aqueous solution. The latest SAMPL9 challenge overview 7 reported on the results of the challenge for 8 cyclodextrin–phenothiazine systems and 13 ligands of the WP6 pillar[6]arene host. It was observed that “methods applying force fields [and molecular dynamics (MD) techniques] achieved in general better correlation with experiments for WP6 [as] opposed to the machine learning and docking models”.…”

“…Nonetheless, MD-based alchemical methodologies 11–13 for absolute dissociation free energy (ADFE) calculations yielded predictions with significant discrepancies, even when using the same force field for the host and guest models. 7,14 According to organizers, the “origins of such differences remain unclear”'.…”

“…The latter choice was dictated by the fact that force field-based methods in SAMPL show particularly severe errors for sulfur-containing compounds. 7…”