IBM Bioinformatics and Pattern Discovery Group

IBM Bioinformatics and Pattern Discovery Group provides pattern discovery and motif-based sequence and structural analysis to support gene identification, sequence annotation, and comparative genomics.

Key Features:

• Gene Identification: Employs the Bio-Dictionary Gene Finder (BDGF), which combines sequence composition statistics with database similarity searches to identify genes among open reading frames (ORFs), leveraging redundant patterns from the natural protein sequence space for high sensitivity and specificity in archaeal and bacterial genomes.
• Protein Annotation: Uses the Bio-Dictionary, a collection of amino acid patterns, with a weighted, position-specific scoring scheme robust to over-representation biases to enable rapid and exhaustive annotation of individual sequences and entire genomes.
• Pattern Discovery: Incorporates the Teiresias algorithm for unsupervised pattern discovery to generate a 1D dictionary of motifs (seqlets) from unaligned ORFs across multiple genomes, covering a substantial portion of input amino-acid sequences to support automated functional annotation and local homology identification.
• Structural Analysis: Aligns structural fragments corresponding to seqlet instances in three dimensions to build a 3D dictionary of structurally conserved motifs and to identify sequences with low RMSD errors for local structure characterization.
• Comprehensive Applications: Supports multiple sequence alignment, gene discovery, protein annotation, detection of structural deviations in amino acid sequences, and pattern discovery in event streams.

Scientific Applications:

• Gene Discovery: Identifies previously unreported genes in archaeal and bacterial genomes.
• Protein Function Prediction: Annotates sequences using functional and structural signals captured by amino-acid pattern dictionaries.
• Comparative Genomics: Facilitates comparison of genomic data across species via motif-based sequence and structural matches.
• Structural Biology: Characterizes local protein structures through 3D motif alignment and RMSD-based assessment of structural conservation.

Methodology:

Uses the Bio-Dictionary Gene Finder (BDGF) combining sequence composition statistics with database similarity searches; applies a weighted, position-specific scoring scheme for Bio-Dictionary annotation; employs the Teiresias algorithm for unsupervised pattern discovery to create 1D seqlet dictionaries from unaligned ORFs; and aligns structural fragments of seqlet instances in three dimensions to construct a 3D dictionary and assess structural conservation via RMSD.