Technical Highlight - June 2013
Short description: The Brotherhood algorithm groups proteins into functional families using indirect sequence similarity.
Until detailed structures can be readily generated for any protein, effective computational methods are necessary to predict protein function. What a protein does can be inferred by sequence similarity to other proteins with known functions. Yet this strategy omits functional relationships characterized by low sequence similarity, as is often the case across large protein superfamilies. Relaxing similarity thresholds can help, but leads to many false functional assignments.
Fiser, Almo and colleagues (PSI NYSGRC and IFN) introduce the Brotherhood algorithm, which uses intermediate sequence analysis to boost the accuracy of detecting related molecules. The Brotherhood intermediate sequence analysis measures the relatedness of two proteins by the degree of overlap in their BLAST-derived sets of similar, or 'intermediate,' proteins. Since the approach can cluster very weakly related proteins, the Brotherhood algorithm retains high specificity by normalizing the number of intermediate sequences by the total number of related sequences.
The authors used the Brotherhood method to classify 561 cell surface or secreted proteins within the immunoglobulin superfamily (IGSF). The IGSF is a diverse superfamily of regulatory proteins that mediate cell adhesion and immunity. At an empirically determined threshold of 45% intermediate protein overlap, Brotherhood largely recapitulates 14 well-curated families, generating only about half the number of singletons as the commonly used programs CD-HIT, which is based on pairwise BLAST similarities, and SCI-PHY, which uses multiple alignments and phylogenomic inferences.
The more inclusive clusters allow new functional predictions, which are highlighted in a family of five characterized nectin and four nectin-like proteins. Brotherhood analysis predicted five new members, including the class-I-restricted T-cell-associated molecule (CRTAM), which it classified as a nectin-like protein.
The researchers used molecular replacement to determine the crystal structure of the Ig-V portion of the CRTAM extracellular domain at 2.3-Å resolution (PDB 3RBG). CRTAM exhibits an antiparallel dimer structure, conserved binding interface residues and a gene structure similar to nectin-like proteins, thus validating its classification. The analysis predicts an unexpected function for CRTAM in mediating homodimer transinteractions.
The Brotherhood classification will help prioritize candidates for structural studies of the IGSF in order to shed light on these important therapeutic targets for cancer and infectious and autoimmune diseases.
R. Rubinstein et al. Functional classification of immune regulatory proteins.
Structure. (11 April 2013). doi:10.1016/j.str.2013.02.022