QUBO-inspired Molecular Fingerprint for Chemical Property Prediction
This work addresses the challenge of fingerprint selection in computational chemistry, offering a novel optimization-based approach that is incremental in nature.
The authors tackled the problem of selecting effective molecular fingerprints for chemical property prediction by generating interaction fingerprints as products of base fingerprints and transforming the search for optimal combinations into a quadratic unconstrained binary optimization problem, achieving improved performance on the QM9 dataset.
Molecular fingerprints are widely used for predicting chemical properties, and selecting appropriate fingerprints is important. We generate new fingerprints based on the assumption that a performance of prediction using a more effective fingerprint is better. We generate effective interaction fingerprints that are the product of multiple base fingerprints. It is difficult to evaluate all combinations of interaction fingerprints because of computational limitations. Against this problem, we transform a problem of searching more effective interaction fingerprints into a quadratic unconstrained binary optimization problem. In this study, we found effective interaction fingerprints using QM9 dataset.