Automated generation of sequence motifs and three dimensional models of proteins and their applications

Theoretical structural models of protein aid to fill the rapid differences between structures and sequences available in the biological databases. In order to sensitively detect structurally equivalent regions in alignments and to identify related sequences we have derived physico-chemical properties based descriptors for amino acids using a large number of properties by applying multi-dimensional scaling. These descriptors were used in a novel motif-detection algorithm incorporated into MASIA (Multiple Alignment Sequence Investigator and Analysis) for automatic identification of conserved regions in proteins families and superfamilies. Physico-chemical profiles created by MASIA were used for scoring alignments, to detect homologus proteins in the twilight-zone and to improve alignment for modeling. We have applied our procedure for data-mining of DNA repair APE related proteins in genomic sequence and structural database. An integrated modeling suite MPACK along with MASIA has been objectively tested and applied to large-scale three-dimensional modeling of proteins in CASP4 and CASP5 competitions. These tools were used for modeling biologically important proteins. We have modeled anti-apoptotic protein Bcl-2 that is regulated by post-translational modification. The model revealed that in Bcl-2 helix 2 rotates to expose hydrophobic residues necessary for the dimerization. Models of immunity/infection related protein human Decay Accelerating Factor (DAF) helped to distinguish the complement and Dr adhesin binding regions. Allergen protein models for Cry j 1 and Jun a 1 were used to identify structural differences in their major IgE epitopes.