Learning Structural Motif Representations for Efficient Protein Structure Search
Authors: Liu et al. Date: 2018
We can compare protein structures in two ways: alignment-based, and alignment-free. One of the most widely-used alignment methods is TMalign, which tries to explicitly superimpose two proteins on each other while minimizing a discrepancy score. The problem with alignment-based methods is that they often do not scale when trying to compare many structures against each other which is the setting in structure retrieval. Alignment-free methods instead encode protein structures into a vector space where comparisons are very quick.
Related work: