Portfolio

Research Focus: Within the realm of cheminformatics, molecular similarity refers to how much in common two molecules have with each other in terms of functional groups, atoms, and chemical bonds. When calculating the similarity value, molecular fingerprints are a popular way to represent these molecules. These fingerprints are essentially strings composed of 1s and 0s with 1 indicating that a feature is present, and 0 for the absence of a feature. Using these fingerprints, a Tanimoto Similarity coefficient can be calculated to quantify the area of overlap between two fingerprints (i.e. having the same byte on the same index). With this coefficient, one can assess how much in common the molecules have with each other on a scale between 0 to 1, with 0 indicating no similarity and 1 for being fully identical. Unfortunately, functions from popular cheminformatic libraries that implement the Tanimoto coefficient only allow comparisons between two singular molecules at a time. This means that if we were to run a full scale molecularity search on a dataset, it would require O(N squared) time complexity where N is the number of molecules, significantly hindering its execution time. The MQ Lab presents a novel method to reduce this horrendous time complexity to a linear trend by developing the iSIMs approach which uses the concept of extended similarity. That is, performing the Tanimoto Similarity index on behalf of the entire dataset as an average by leveraging an attribute called linear sums. With this brilliant implementation, it takes significantly less time to go through the dataset and cluster out molecules based on into their respective groups based on a similarity index.