Drug binding and docking: where quantum helps and where classical wins
The central number in drug design is binding free energy: how tightly a candidate molecule holds onto its target. An error of 1.36 kcal/mol changes the predicted binding constant tenfold at room temperature. Computational methods now routinely predict relative binding energies to about 1 kcal/mol for well-behaved systems. Where does the remaining error come from? If it came mostly from inaccurate electronic structure, quantum computers would help broadly. It mostly does not.
On the plate: the error budget of a binding prediction.
How binding is computed
Docking places a molecule in the binding site and scores the pose with a fast function, screening millions of compounds. Free-energy perturbation (FEP) is slower and more rigorous: it uses molecular dynamics with a classical force field to “alchemically” transform one ligand into another, and measures the free-energy difference from many sampled configurations. Modern FEP workflows reach mean errors around 1 kcal/mol on benchmark sets. The free energy relates to the dissociation constant by ; at 298 K, kcal/mol per factor of ten. Established
Where the error comes from
Set the contributions, in kcal/mol: conformational sampling 0.8, force field 0.7, solvation and protonation 0.5, electronic structure 0.3. Independent errors add in quadrature. For a typical non-covalent drug, the defaults reflect the field’s experience: sampling and force fields dominate. Setting electronic error to zero, which is what a perfect quantum calculation would do, barely changes the total. The picture changes for covalent inhibitors, which form a bond to the protein, and for drugs targeting metalloproteins: there the electronic term can dominate, and QM/MM with an accurate quantum core (Chapter 160) is needed. Nicholas Blunt and colleagues estimated in 2022 that even the covalent drug ibrutinib’s reaction with its kinase target would need an error-corrected quantum computer to beat classical methods. Established for the budget; Debated for the quantum role.
The critique
Raffaele Santagati and colleagues, from Boehringer Ingelheim, Google and elsewhere, argued in a 2024 Nature Physics perspective that if quantum advantage is limited to strongly correlated electronic structure, its direct impact on mainstream drug design may be limited, since most drug targets are not of that kind. They saw the clearest value in specific cases (metalloenzymes, covalent chemistry, photochemistry) and in generating accurate reference data to train faster classical and machine-learning models. That is the honest current picture. Debated, leaning classical for most targets.
What you now have
Docking and free-energy perturbation, the link between free energy and binding constants, an error budget for binding predictions, why sampling and force fields usually dominate, where electronic structure matters (covalent, metal-binding drugs), and the Santagati et al. outlook.
· · ·
Three questions to carry with you.
And one to work out. With kcal/mol at 298 K, what is ?
Sources: L. Wang et al., J. Am. Chem. Soc. 137, 2695 (2015) (FEP+). N. S. Blunt et al., J. Chem. Theory Comput. 18, 7001 (2022). R. Santagati et al., Nat. Phys. 20, 549 (2024). The default error contributions are illustrative, not measured values.