Yea, I mean basically there are a variety of techniques for encoding molecules as vectors. Morgan/circular fingerprints (cf https://darkomedin-datascience.medium.com/data-science-for-d... as a quick google search) are one way of encoding to bit vectors, and the Tanimoto similarity is a well validated measure of similarity (because of the bit vectors, cf https://en.wikipedia.org/wiki/Chemical_similarity).

Virtual screens of molecules frequently use steps to filter out molecules, one of which is via chemical similarity.