What is a good AutoDock Vina score?

You ran a dock, you got back a number like −8.4, and now you want to know one thing: is that good? It is the most common question about docking, and the honest answer is that a Vina score is a predicted binding energy, not a measured one — so “good” is less about clearing a fixed threshold and more about what the number is telling you relative to everything else. Here is how to read it without fooling yourself.

What the number actually is

AutoDock Vina reports its score in kcal/mol, as an estimate of the binding free energy between your ligand and the receptor. It is negative because binding releases energy, and a more negative value means a stronger predicted interaction. The scoring function behind it — introduced by Trott and Olson in 2010 — is an empirical mix of steric, hydrophobic, and hydrogen-bonding terms tuned to reproduce known binding data. It is fast and surprisingly good at finding the right pose, but it is an approximation of affinity, not a measurement of it.

Rough ranges, for orientation only

With that caveat loud and clear, here are the ballparks you will see for a drug-like molecule docked into a well-formed pocket:

• Weaker than −6: usually a poor fit, a shallow or wrong pocket, or a fragment-sized molecule.
• −6 to −7: modest — a hint of binding, worth a second look but not exciting on its own.
• −7 to −9: the typical range for a decent drug-like hit in a real binding site. Most useful results live here.
• −9 to −11: strong — but start asking why it is so negative.
• More negative than −11: unusually strong, and very often a red flag rather than a win (see below).

Treat those bands as orientation, not a grading rubric. The “good” cutoff shifts with the target, the pocket, and the size of your molecule.

Why a single score lies to you

Three things make any one Vina number untrustworthy on its own:

• Run-to-run noise. Vina’s search is stochastic, with a documented noise floor of roughly ±1 kcal/molat default settings. A difference smaller than that can flip sign if you simply re-run the job — it is not a real difference.
• It is not a measured K_d. The classic Warren et al. (2006) assessment ran ten docking programs across eight targets and found that scoring functions place poses reasonably but predict absolute affinity poorly. Reading −9.2 as a nanomolar binder is exactly the mistake the field has warned about for two decades.
• Scores do not compare across targets. A −8 on one protein and a −8 on another are not the same thing. Different pockets, box sizes, and receptor preparations shift the baseline.

The big trap: size and grease inflate the score

Vina’s score grows, roughly, with the number of favorable contacts a molecule can make — so bigger, greasier molecules rack up more negative scores almost automatically, whether or not they are better drugs. A 700 Da, LogP 8 compound scoring −10.5 is usuallyworse than a 320 Da, LogP 2 compound scoring −8: the first is a developability dead end, the second is a real lead. The raw score hides this completely.

The fix is to normalize for size with ligand efficiency (LE = −score ÷ heavy atoms) and for lipophilicity with LLE and LELP. A −10 spread over 45 heavy atoms (LE 0.22) is mediocre; a −8 over 22 heavy atoms (LE 0.36) is excellent. We break the math down in Ligand efficiency metrics: LE, LLE, and LELP explained, and Liganx now shows all three right next to the Vina score so a size-inflated “hit” is obvious at a glance.

How to use a Vina score properly

Scoring functions are mediocre at absolute affinity but genuinely useful for ranking related things scored the same way. So the number worth trusting is almost never one score — it is a difference: this analog versus that analog on the same receptor, or the same ligand on a wild-type versus a mutant protein. That wild-type-minus-mutant difference (a ΔΔ) is far more robust than either absolute value, because the shared errors cancel. We make the full argument in Why ΔΔ beats absolute docking scores.

Try it yourself

Liganx is molecular docking online — free and browser-based, no install. Open Studio, dock your compounds, and read the Vina score the right way: check the ligand-efficiency panel next to it, and compare against a wild-type or a reference rather than trusting a lone number. A good Vina score is not a magic threshold — it is an efficient, reproducible one that beats the right comparison.

Primary sources

• Trott O, Olson AJ. AutoDock Vina: Improving the Speed and Accuracy of Docking with a New Scoring Function, Efficient Optimization, and Multithreading. J Comput Chem 31, 455–461 (2010). doi:10.1002/jcc.21334
• Warren GL, et al. A Critical Assessment of Docking Programs and Scoring Functions. J Med Chem 49, 5912–5931 (2006). doi:10.1021/jm050362n
• Hopkins AL, Groom CR, Alex A. Ligand Efficiency: a Useful Metric for Lead Selection. Drug Discov Today 9, 430–431 (2004). doi:10.1016/S1359-6446(04)03069-7