Prediction of protein function from theoretical models

Abstract

Explicit 3D models can be obtained by comparative protein modelling, a mature and predictable technique, fragment assembly ab initio methods for smaller novel or unrecognisable folds and contact-based methods for large protein families. Each modelling method has limitations in model accuracy, which vary further according to the characteristics of the target: as a result, the performance of structure-based function prediction algorithms applied to models is variable. Nevertheless, with care, a wide variety of structure-based methods can be productively applied to protein models, frequently facilitating the planning and interpretation of experimental results. This chapter will first survey the literature on applicability of structure-based methods specifically to models, before discussing a selection of examples in more detail.