Core Technology

EigenBio's immunome mapping is built around a novel applied mathematical platform called uTOPE™ analysis, a systems biology approach which:

      • Enables detailed immunologic analysis of any protein
      • Provides powerful statistical analyses to compare proteins
      • Integrates components of systemic biological responses.

The multiple modules of uTOPE™ analysis are driven off peptide descriptor algorithms. Principal component analysis was used to generate descriptive algorithms for amino acids based on a wide variety of physical and chemical characteristics. Neural nets and partial least squares regression was used to train the algorithms for short peptides (typically 9-mer and 15-mers) by comparison to experimental results (training sets) for MHC binding, B cell and antibody binding and cathepsin cleavage. The resulting uTOPE™ predictions were validated against experimental results.

uTOPE™ analysis is a very versatile approach with multiple modules; the principal among these are:

      • Predicted MHC-I and MHC-II binding
      • Predicted B-cell or antibody binding epitopes
      • Cleavage by cathepsin B, L and S
      • Protein topology
      • T-cell recognition motifs and predicted Tregulatory cell binding epitopes

uTOPE™ analysis of a protein generates a database of binding by allele and cleavage data for that protein, and a number of graphical outputs are derived from these. Click here to view examples of graphical outputs. Secondary analysis from uTOPE™ databases includes simulations of the impact of amino acid changes, cross correlations of proteins and multiple protein clustering analyses.

By integrating multiple layers of information uTOPE™ analysis builds up a deeper understanding of the interaction of different facets of the immune response, much as a geographic information system can be used to describe a landscape. Particular advantages of the uTOPE™ system are that the systems biology approach used has rigorous mathematical underpinnings based on principal component analysis, neural networks, partial least squares regression, and clustering algorithms that reduce the dimensionality of multivariate data. Unlike legacy approaches it can be applied to whole proteins and proteomes and generates high data density enabling higher order pattern recognition. By converting peptides to mathematical algorithms a wide range of analytical tools become available. Prediction equations generate a mean as well as a standard deviation of a binding affinity, impossible through the use of alphabetic motifs. uTOPE™ predictions acknowledge that binding of any peptide is competitive and apply statistical standardization between alleles within proteins (or within proteomes).