INTREPID Functional Residue Prediction Server Help

Home | PhyloFacts | BPG Tools | FAQ | Help

Details on the INTREPID algorithm (Sankararaman and Sjölander, Bioinformatics 2008) for functional residue prediction can be found here.
A step-by-step usage guide of the server can be found in our tutorial.
Follow this link to the INTREPID Functional Residue Prediction Server.

Overview

Not all the positions in a protein sequence are critical for protein function. Some residues, if mutated, may have little effect on the protein's function, while mutating other residues might completely disrupt protein function. Determining which positions are critical for protein function is a major challenge in bioinformatics.

Residues critical for function are likely to be conserved through evolution, whereas less important positions are more likely to vary across members of the family. We have developed algorithms that reconstruct and analyze the evolutionary history of a protein family to determine the specific positions that are critical for maintaining molecular function across the family as well as within subfamilies. These algorithms are provided via a web-based interface.

This phylogenomic approach to functional residue prediction has many applications in different functional contexts, including predicting enzyme active sites, predicting residues that are crucial for supporting interactions with other molecules, and predicting positions responsible for stabilizing the overall protein fold.

Pipeline

Figure 1: Overview of the Functional Residue Prediction server pipeline. Prediction of functional residues using the INTREPID algorithm performs a number of complex tasks. The computation time required for the analysis is dominated by the modified PhyloBuilder pipeline (A). Retrieval of homologous sequences from the UniProt database typically takes ~10 minutes or more. Constructing a multiple sequence alignment can take anywhere from a few seconds to ~20 minutes, depending on how many homologous sequences were retrieved. Inferring the phylogenetic tree may take 30 minutes or more, again depending on how many sequences are in the alignment. Finally, scoring the PDB database for homologous 3D structures can take ~10 minutes. Once these data have been computed, functional residue prediction using INTREPID typically requires only a few minutes of computation time.

A The user submits a protein sequence to the Functional Residue Prediction server using a web browser interface.

B We use a modified version of the PhyloBuilder pipeline (Glanville et al, Nucleic Acids Research 2007) to gather homologous sequences, construct a multiple sequence alignment and infer a phylogenetic tree. Details of the clustering and alignment methods used are available in Sankararaman and Sjölander, Bioinformatics 2008.

C The results of PhyloBuilder include a phylogenetic tree describing the evolutionary history of all sequences in the protein family, a multiple sequence alignment (MSA), and a list of homologous 3D structures. These data are displayed in a PhyloFacts book.

D The phylogenetic tree and multiple sequence alignment, are used as inputs to the INTREPID algorithm to predict functional residues. INTREPID uses the branching structure of the phylogeny to analyze patterns of sequence conservation in the MSA.

E The Functional Residue Prediction server displays predicted residues on a user-selected 3D structure, highlighting spatial patterns of sequence conservation at various significance thresholds. The raw prediction data are also available for download and further analysis.

Interested in learning how to use the pipeline? Please read our Tutorial.