>>START THE APPLICATION<<

Aims and Objectives

We aim to develop a distributed computational pipeline which makes a large number of immunoinformat tools available through one front-end. This process has begun with the Antigenic Peptide Prediction Pipeline (APPP) which combines a variety of peptide-binding scoring methods and proteasomal cleavage scoring methods to assess the potential antigenic peptides from a protein sequence.

ImmunologyGrid

The aim of the ImmunologyGrid project is to to integrate a series of programs and methods which together will form an easy-to-use Grid-enabled tool for in silico prediction of binding affinities between MHC molecules and associated peptides. Improved knowledge of peptide binding has far-reaching benefits in many areas of molecular medicine and immunology, such as histocompatibility (tissue transfer), vaccine design, autoimmunity and allergy research.

Antigenic Peptide Prediction Pipeline

Background and Methods

Peptide Processing & Presentation
The form in which protein antigens are recognised by the cellular immune system is as peptides bound by MHC molecules. Antigen processing and presentation - the fragmentation of proteins to peptides, the subsequent interaction of those peptides with MHC molecules and their display at the cell surface for recognition by T cells - does not intrinsically discriminate between self and non-self proteins. Instead, different antigen processing pathways discrimi-nate the location of the antigen with respect to the cell. These pathways are called class I and class II MHC (See Figure 1).

Figure 1. Antigen processing pathways

Peptides displayed by MHC class I molecules are products of the degradation of protein by proteasomes, cylindrical arrays of prote-olytic enzymes with their active sites towards the centre of the cylinder. Both pathogen proteins and self cell proteins can be com-plexed with ubiquitin to target them to the proteasome for process-ing. Two proteases encoded in the MHC Class II region LMP2 and LMP7) and a third subunit not encoded in MHC is produced in response to interferon, which is synthesised in response to virus infection. These inducible proteases replace constitutive proteases in the proteasome and produce peptides with basic and hydropho-bic carboxyl terminal residues preferred as anchor residues in class I peptide-binding sites and for transport from the cytosol into the ER.

Immunoinformatic Components

Immunoproteasome Prediction
Proteasomal cleavage of proteins is the first step in the processing of most antigenic peptides that are presented to cytotoxic T-cells. For this reason we have included a tool which estimates the immunoproteasome cleavage which are based on the weight matrices used in the propred-I server. These predictions are translated into a cleavage score which estimates the cleavage potential of an amino acid pair. We have also developed a second immunoproteasome cleavage potential prediction tool based on the data from the work of Altuvia and Martgalit's. This work involved a rigorous analysis of the residues at the termini and flanking regions of 286 naturally processed peptides eluted from MHC class I molecules. Their results showed that both the C terminus and its immediate flanking position possess significant signals. This component 'windows' through a protein sequence evaluating every peptide for its potential of being cleaved by the immunopro-teasome. See Figure 2 for a diagram showing the windowing of a protein string.

Figure 2. Method of predicting the immunoproteasomal cleavage of a protein

Peptide Binding prediction
These components utilise a variety of matrix based scoring methods for evaluating the binding affinity of peptides to a MHC allele. Four different methods are used: CombiPred (Shaban, R PhD Thesis). BIMAS is a server which allows users to locate and rank 8-mer, 9-mer, or 10-mer peptides that contain peptide-binding motifs for HLA class I molecules. MMBPred is a server for the prediction of mutated high affinity and promiscuous MHC class-I binding peptides from protein sequence. SYFPEITHI is a profile based prediction scoring method for the prediction of MHC class I & II peptide binding. Currently the APPP only utilises MHC class I binding predictions. This is due to the greater availability of scoring methods for class I.

Consensus Finder
This component, specially designed for this pipeline, evaluates the outputs from all the different components used throughout the workflow and discovers any consensus that may exist between them. It then returns the results listing the peptides which have been predicted as antigenic by all of the components. Work is ongoing to expand this tool to allow a more realistic inference of consensus between the various components.

Workflow

The steps that form the process of Antigenic Peptide Prediction can be described as a workflow. A workflow describes the control and data flow through a series of processes. In the case of Computational Grids, the processes are generally represented as software components. The description of the order in which these components operate and the way data is passed between them is known as the workflow description. Business Process Execution Language (BPEL) is a language for specifying workflow descriptions that is standardised through the OASIS standards body. BPEL was formed out of two earlier workflow languages, IBM's Web Services Flow Language (WSFL) and Microsoft's XLang. BPEL has been designed specifically to operate with Web Services as the component model.

The componentised model that enables a workflow of services to be built is known as a Service Oriented Architecture (SOA). Web Services provide one means of implementing a SOA. The Web Services paradigm is based on a series of standards that specify how services may be advertised, discovered and how they communicate with each other. UDDI is used for publishing and discovering services. Web Services Description Language (WSDL) is used to specify a service -- its operations, data types and bindings -- while Simple Object Access Protocol (SOAP) is used as the message format for communication between services and their clients. SOAP is transport independent and may be bound to different transport mechanisms such as HTTP or SMTP. The utilisation of Web Services allows us to specify our pipeline in an abstract manner without the need to bind services to specific resources prior to runtime, thus providing greater flexibility.

The Antigenic Peptide Prediction Pipeline (APPP) is made up of six web services deployed on a J2EE compliant application server at LeSC, Imperial College. Figure 3 shows how these services are connected together in a workflow to form the APPP.

Figure 3. Design of the antigenic peptide prediction pipeline

Run the Application

A web-based interface to our pipeline has been designed using the OpenLaszlo platform. OpenLaszlo allows developers to create applications with the rich user interface capabilities of desktop client software and the instantaneous no-download Web deployment of HTML. This interface can be found here.

Status and Availability

The project is currently under development.

Contacts

For further details contact info@ImmunologyGrid.org, or see Prof David Moss's project page , Dr Barry R Smith's project page or Mark Halling-Brown's project page.