Membrane proteins (MPs) have unique structural features to interact with both phospholipid cell membranes and the aqueous environment. Almost 20-30% of genes in every genome encode MPs and these proteins perform several functions including transporters, enzymes, receptors and communicators in living organisms. Notably, in the human proteome 60% of drug targets are MPs. Mutations in MPs may alter their folding, stability, functions, and some of them lead to diseases such as cancers, cystic fibrosis, etc. In this work, we have developed computational databases and tools for exploring the role of mutations in membrane protein stability and diseases.

We have developed a specific and reliable database MutHTP, which contains information about disease-causing and neutral mutations on membrane proteins. MutHTP provides several sequence, structure, disease and membrane protein-specific features for each mutation. We also provided the various search and display options to retrieve the data [1]. This database can be accessible at https://www.iitm.ac.in/bioinfo/MutHTP/.

We effectively used the MutHTP database to explore disease-causing and neutral mutations in different types of membrane proteins and diseases. In the human membrane proteome, we observed that negatively charged to positively charged/polar and nonpolar to nonpolar changes are dominant in disease-causing and neutral mutations, respectively. We also investigated mutations in 14 different disease classes and found that Arg residue is prevalent in most of them [2]. Further, we have constructed a sequence-based tool (Pred-MutHTP) to classify the disease-causing and neutral mutations with respect to different topological regions. The topology-specific (cytosol-, membrane-, and extracellular) machine learning models showed an average accuracy of 80% on the respective test datasets. Moreover, we compared the performance of Pred-MutHTP with existing methods in the literature and observed an improvement of 4-11% in balanced accuracy [3]. The tool is available at https://www.iitm.ac.in/bioinfo/PredMutHTP/.

To understand the role of mutations on membrane protein stability, we developed the MPTherm database, which has the thermodynamic/stability data of membrane protein and their mutants. MPTherm database provides protein sequence and structural information, membrane topology, experimental conditions, thermodynamic parameters such as melting temperature, free energy, enthalpy, etc., and literature information for each entry [4]. The database is accessible at https://www.iitm.ac.in/bioinfo/mptherm/. This is a potential resource to design stable mutants on membrane proteins for different application purposes.

Utilizing the MPTherm database, we have developed a suite of topological specific computational methods to predict the change in thermal stability upon missense mutations. We observed that the factors such as changes in hydrophobicity and the average number of contacts between Cα atoms are essential determinants of the stability change induced by mutants in transmembrane regions. The topology-specific multiple linear regression models showed a correlation and MAE of 0.74 and 3.2°C, respectively, between experimental and predicted stabilities on 10-fold cross-validation. It could also successfully distinguish between stabilizing and destabilizing mutants at an accuracy of 77% [5]. These methods are available as a web server at https://web.iitm.ac.in/bioinfo2/mpthermpred/.

In essence, the present study would be useful to explore the relationship between disease-causing mutations and their stability. Besides, it will help to develop mutation-specific drug design strategies for different diseases.