Combinatorial pattern discovery in biological sequences: the TEIRESIAS algorithm. [Erratum: 1998, v. 14 (2), p. 229.]

• Published in: Bioinformatics (Oxford, England) Vol. 14; no. 1; pp. 55 - 67
• Main Authors: Rigoutsos, I, Floratos, A
• Format: Journal Article
• Language: English
• Published: Oxford Oxford University Press 1998
• Subjects: Algorithms; Amino Acid Sequence; amino acid sequence motifs; amino acid sequences; Animals; Biological and medical sciences; Conserved Sequence; Evaluation Studies as Topic; Fundamental and applied biological sciences. Psychology; General aspects; Histones - chemistry; Humans; Leghemoglobin - chemistry; Mathematics in biology. Statistical analysis. Models. Metrology. Data processing in biology (general aspects); Molecular Sequence Data; Pattern Recognition, Automated; patterns; Sequence Alignment - methods; Sequence Analysis; Structural unit; Investigation method; Nucleotide sequence; Computerized processing; Sequence alignment; Algorithm; Nucleic acid; Comparative study; Protein; Aminoacid sequence
• ISSN: 1367-4803; 1367-4811
• DOI: 10.1093/bioinformatics/14.1.55

MOTIVATION: The discovery of motifs in biological sequences is an important problem. RESULTS: This paper presents a new algorithm for the discovery of rigid patterns (motifs) in biological sequences. Our method is combinatorial in nature and able to produce all patterns that appear in at least a (user-defined) minimum number of sequences, yet it manages to be very efficient by avoiding the enumeration of the entire pattern space. Furthermore, the reported patterns are maximal: any reported pattern cannot be made more specific and still keep on appearing at the exact same positions within the input sequences. The effectiveness of the proposed approach is showcased on a number of test cases which aim to: (i) validate the approach through the discovery of previously reported patterns; (ii) demonstrate the capability to identify automatically highly selective patterns particular to the sequences under consideration. Finally, experimental analysis indicates that the algorithm is output sensitive, i.e. its running time is quasi-linear to the size of the generated output.