Tautomers are structural isomers of a given molecule that readily interconvert. The most common form of tautomerism is prototopic tautomerism, which occur when one or more protons migrates around a molecule.
Understanding which tautomer(s) will predominate in solution is critically important for predicting and interpreting the behavior of functional molecules. In medicinal chemistry, tautomerism is ubiquitous—many motifs common to drugs are prone to tautomerism, and accurate structure-based drug design and analysis of structure–activity relationships requires accurate identification of low-energy tautomers. Since different tautomers often have dramatically different shapes, polarity, and biophysical properties, neglecting tautomers can lead to misleading trends or nonsensical activity cliffs in aggregated data. In addition, faulty tautomer prediction will poison all downstream calculations with the incorrect input structure!
Since measuring the tautomer ratio of a given compound experimentally can be time-consuming, it's standard to predict tautomer ratios in silico usign computational techniques. While simple instances of tautomerism can often be resolved by pattern-matching or analogy to known tautomer pairs, more complex tautomeric equilibria require consideration of three-dimensional conformation and the delicate stereoelectronic effects particular to each molecule. Computations based on quantum chemistry can handle these factors and are thus state-of-the-art for tautomer prediction, but are often slow and somewhat involved, making their application to large datasets challenging.
Rowan uses machine-learned interatomic potentials to accelerate the accurate computation of tautomer ratios. Using our efficient in-house conformational search procedures, we sort through all potential prototopic tautomers for each molecule, generate a conformational ensemble, and then score this ensemble using machine learning-based simulations corrected with modern implicit solvent methods. Our methodology is almost as accurate as quantum chemistry while running in a mere fraction of the time: see the blog post for more details!
Here's the result of a tautomer search in Rowan:
Rowan automatically computes the Boltzmann weight associated with each tautomer and sorts the ensemble, making it easy to see what the tautomer ratio will be in solution. Different tautomers and their corresponding geometry can be visualized by clicking on their corresponding row, allowing for quick inspection of the overall ensemble, and each of these structures can be used as the input to a subsequent Rowan calculation.
