Automatically detecting crucial protein features

By Mike Nagle

- Last updated on GMT

A new tool can automatically analyse a protein's structure to
detect residues crucial to its function, allowing pharma firms to
design better drugs.

The computer programme, which can identify which amino acid residues in a protein are critical to both its structure and its function, was developed by Dr Gabriel del Río at the Universidad Nacional Autonoma de Mexico along with Professor Dale Bedesen the Buck Institute in the USA. Data on which residues are the most important in binding a small molecule to a protein can then give scientists a clearer idea which molecules will interact with the protein. Not only can this help decipher the protein's function, it could also enable researchers to more easily design molecules that bind to the protein better than its natural partner and either mimic that substrate's effect or block it - leading to better drugs. Other tools to identify critical residues have been developed; however, most of them analyse the 2-D sequence of amino acids that form the protein despite the fact that it is the 3-D structure that determines a protein's function. By analysing a collection of related sequences, scientists can see which amino acids are always present in the same place and so designate those residues as critical. However, this becomes difficult if there are not enough related sequences. At least 25 per cent of proteins with a known structure do not show significant sequence similarity with other proteins, claim del Río and his colleagues in an article presenting their research in the 9 May issue of the journal, Public Library of Science One​. Therefore, structure is often the only information to go on. The authors also say that programmes that do take structure into account are few and far between or use sequence analysis as part of their approach. When a protein is inputted into the new tool, the programme's first step is to build a network based on the structure. Any residues within 5 angstroms (or a distance set by the user) of each other are considered neighbours and paired together. Also, four chemical property filters were applied to amino acids in order to generate and test different networks. Once the network is generated, the algorithm then connects every pair of residues through the shortest path possible. The number of times each residue is traversed is then counted to generate a 'dynamic connectivity' number. This figure can then be used to calculate which residues are critical. "We are interested in identifying critical residues with the highest specificity and lowest error values, even if the sensitivity is low,"​ said the authors. The sensitivity can be improved if more than one alternative structure of a given protein is inputted into the tool. However, the research suggests that analysing more and more structures does not necessarily lead to better results; instead, it is the diversity of the inputted structures that correlates to improved predictions. This also ties in with the fact that, in reality, protein structure is dynamic with various structures contributing to a given protein's function. The tool is based on the Java programming language, which has multitasking capabilities that helps the tool to cope with analysing multiple structures. It also means the programme can be run on multiple computers or even online​. A sample output can be seen here​.

Related topics: Preclinical Research

Related news