homology model

A homology model is an educated guess at a protein's three-dimensional shape, built by borrowing the known structure of a relative. If you know the shape of a cousin protein and your target shares much of its sequence, you can thread your target's amino acids onto that template, much like tailoring a new suit from a pattern that fit someone of similar build.

The method rests on a biological fact: proteins with similar sequences usually fold into similar shapes. The steps are to find a suitable experimental template, align the target sequence to it, copy the template's backbone where they correspond, then model the differing loops and side chains and refine the result. The closer the sequence identity between target and template, the more trustworthy the model.

Homology models let medicinal chemists do structure-based design and docking for targets that have never been crystallized, which is invaluable for many receptors. The honest caveat is that accuracy degrades sharply at low sequence identity, and the binding pocket and flexible loops, precisely the parts a chemist cares about most, are often the least reliable. Modern deep-learning structure predictors have raised the quality bar, but any predicted structure should be validated before betting a program on it.