In the chemoinformatics area, it is important to describe molecular similarity. Merit of fingerprint bit vector based similarity calculation is speed I think. But sometime ECFP4 or any other related methods do not sense of chemst feeling. By the way graph based similarity like a MCS is useful but calculation cost is high. You know, gWT ( graph indexing wavelet tree ) works very first for similarity search.
And also new version of RDKit implemented Atom-Atom Path based similarity algorithm. I’ll introduce the FP soon…
Recently I found very attractive report.
A Novel Graph-based Approach for Determining Molecular Similarity
First author is researcher of 1QBit. 1QBit developer tools and end-to-end expertise in quantum software are built to solve the world’s most demanding computational challenges( from HP ).
;-) The article solve the molecular similarity based prediction used “D-Wave systems“.
D-wave system is the quantum computer based on quantum annealing (QA). QA is a metaheuristic for finding global minimum of given objective function by a process using quantum fluctuations.
They represent molecule as graph. The graph has many information, i.e. kind of atoms, bonds, formal charge, position, geometrical center of ring. In the article, the author did not describe the details of the way to molecule to graph.
Finally, they performed case study “prediction of mutagenicity”. And they compared with graph similarity based method and MACCS fingerprint based method. And graph based method showed good performance.
The results was very exiting for me. Because the result indicated that QA is useful for chemoinformatics.
I want to learn about QA more and more.
Keep watching QA!