Hidden Markov Models for detection of intrinsically disordered regions of proteins

The aim of this research is to build a collection of Hidden Markov Models (HMMs) that can accurately identify intrinsically disordered protein (IDP) domains as defined in the manually curated databases. IDP domains are characterized by the lack of a fixed three dimensional structure and therefore their evolution is not constrained by the need of preserving a well defined topology. This is reflected by a lower sequence conservation compared to globular domains. IDP domains are essential to cellular processes such as alternative splicing, multivalent and transient binding, cellular signalling and regulation. The project focuses on modelling the subset of IDP domains that retain some sequence conservation, as those that bind a partner molecule. The work starts from building MSA of the of IDP domains as provided by the MobiDB database, followed by the creation of the corresponding HMMs. The accuracy of the resulting HMMs is validated using benchmark datasets and compared with the existing HMM databases. The development of HMMs for disordered domains will improve protein sequence annotation accuracy and facilitate a more comprehensive analysis of these regions. This work will contribute to the advancement of techniques for predicting protein disorder and facilitate the study of disordered regions in biological systems. Finally, the HMMs will be integrated into widely-used protein sequence analysis resources like Pfam and InterPro.