{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,2]],"date-time":"2026-04-02T18:51:59Z","timestamp":1775155919729,"version":"3.50.1"},"reference-count":35,"publisher":"Oxford University Press (OUP)","issue":"14","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2007,7,15]]},"abstract":"<jats:title>Abstract<\/jats:title><jats:p>Motivation: As more genomes are sequenced, the demand for fast gene classification techniques is increasing. To analyze a newly sequenced genome, first the genes are identified and translated into amino acid sequences which are then classified into structural or functional classes. The best-performing protein classification methods are based on protein homology detection using sequence alignment methods. Alignment methods have recently been enhanced by discriminative methods like support vector machines (SVMs) as well as by position-specific scoring matrices (PSSM) as obtained from PSI-BLAST.<\/jats:p><jats:p>However, alignment methods are time consuming if a new sequence must be compared to many known sequences\u2014the same holds for SVMs. Even more time consuming is to construct a PSSM for the new sequence. The best-performing methods would take about 25 days on present-day computers to classify the sequences of a new genome (20 000 genes) as belonging to just one specific class\u2014however, there are hundreds of classes.<\/jats:p><jats:p>Another shortcoming of alignment algorithms is that they do not build a model of the positive class but measure the mutual distance between sequences or profiles. Only multiple alignments and hidden Markov models are popular classification methods which build a model of the positive class but they show low classification performance. The advantage of a model is that it can be analyzed for chemical properties common to the class members to obtain new insights into protein function and structure.<\/jats:p><jats:p>We propose a fast model-based recurrent neural network for protein homology detection, the \u2018Long Short-Term Memory\u2019 (LSTM). LSTM automatically extracts indicative patterns for the positive class, but in contrast to profile methods it also extracts negative patterns and uses correlations between all detected patterns for classification. LSTM is capable to automatically extract useful local and global sequence statistics like hydrophobicity, polarity, volume, polarizability and combine them with a pattern. These properties make LSTM complementary to alignment-based approaches as it does not use predefined similarity measures like BLOSUM or PAM matrices.<\/jats:p><jats:p>Results: We have applied LSTM to a well known benchmark for remote protein homology detection, where a protein must be classified as belonging to a SCOP superfamily. LSTM reaches state-of-the-art classification performance but is considerably faster for classification than other approaches with comparable classification performance. LSTM is five orders of magnitude faster than methods which perform slightly better in classification and two orders of magnitude faster than the fastest SVM-based approaches (which, however, have lower classification performance than LSTM). Only PSI-BLAST and HMM-based methods show comparable time complexity as LSTM, but they cannot compete with LSTM in classification performance.<\/jats:p><jats:p>To test the modeling capabilities of LSTM, we applied LSTM to PROSITE classes and interpreted the extracted patterns. In 8 out of 15 classes, LSTM automatically extracted the PROSITE motif. In the remaining 7 cases alternative motifs are generated which give better classification results on average than the PROSITE motifs.<\/jats:p><jats:p>Availability: The LSTM algorithm is available from http:\/\/www.bioinf.jku.at\/software\/LSTM_protein\/<\/jats:p><jats:p>Contact: \u00a0hochreit@bioinf.jku.at<\/jats:p>","DOI":"10.1093\/bioinformatics\/btm247","type":"journal-article","created":{"date-parts":[[2007,5,9]],"date-time":"2007-05-09T02:24:29Z","timestamp":1178677469000},"page":"1728-1736","source":"Crossref","is-referenced-by-count":104,"title":["Fast model-based protein homology detection without alignment"],"prefix":"10.1093","volume":"23","author":[{"given":"Sepp","family":"Hochreiter","sequence":"first","affiliation":[{"name":"1 Institute of Bioinformatics, Johannes Kepler Universit\u00e4t Linz 4040 Linz, Austria and 2Department of Electrical Engineering and Computer Science, Technische Universit\u00e4t Berlin and Bernstein Center for Computational Neuroscience, 10587 Berlin, Germany"}]},{"given":"Martin","family":"Heusel","sequence":"additional","affiliation":[{"name":"1 Institute of Bioinformatics, Johannes Kepler Universit\u00e4t Linz 4040 Linz, Austria and 2Department of Electrical Engineering and Computer Science, Technische Universit\u00e4t Berlin and Bernstein Center for Computational Neuroscience, 10587 Berlin, Germany"}]},{"given":"Klaus","family":"Obermayer","sequence":"additional","affiliation":[{"name":"1 Institute of Bioinformatics, Johannes Kepler Universit\u00e4t Linz 4040 Linz, Austria and 2Department of Electrical Engineering and Computer Science, Technische Universit\u00e4t Berlin and Bernstein Center for Computational Neuroscience, 10587 Berlin, Germany"}]}],"member":"286","published-online":{"date-parts":[[2007,5,11]]},"reference":[{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"403","DOI":"10.1016\/S0022-2836(05)80360-2","article-title":"Basic local alignment search tool","volume":"215","author":"Altschul","year":"1990","journal-title":"J. Mol. Biol"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"3389","DOI":"10.1093\/nar\/25.17.3389","article-title":"Gapped BLAST and PSI-BLAST: a new generation of protein database search programs","volume":"25","author":"Altschul","year":"1997","journal-title":"Nucleic Acids Res"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"189","DOI":"10.1093\/nar\/24.1.189","article-title":"The PROSITE database, its status in 1995","volume":"24","author":"Bairoch","year":"1999","journal-title":"Nucleic Acids Res"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"937","DOI":"10.1093\/bioinformatics\/15.11.937","article-title":"Exploiting the past and the future in protein secondary structure prediction","volume":"15","author":"Baldi","year":"1999","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"i75","DOI":"10.1093\/bioinformatics\/bti1004","article-title":"Three-stage prediction of protein beta-sheets by neural networks, alignments, and graph algorithms","volume":"21","author":"Cheng","year":"2005","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"349","DOI":"10.1093\/bioinformatics\/17.4.349","article-title":"Multi-class protein fold recognition using support vector machines and neural networks","volume":"17","author":"Ding","year":"2001","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"285","DOI":"10.1093\/bioinformatics\/bti801","article-title":"Application of latent semantic analysis to protein remote homology detection","volume":"22","author":"Dong","year":"2006","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"1437","DOI":"10.1016\/S0022-2836(02)01470-5","article-title":"A comprehensive view on proteasomal sequences: implications for the evolution of the proteasome","volume":"326","author":"Gille","year":"2003","journal-title":"J. Mol. Biol"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"4355","DOI":"10.1073\/pnas.84.13.4355","article-title":"Profile analysis: detection of distantly related proteins","volume":"84","author":"Gribskov","year":"1987","journal-title":"Proc. Natl Acad. Sci."},{"key":"2023041105225738200_","first-page":"94","article-title":"Family-based homology detection via pairwise sequence comparison","author":"Grundy","year":"1998"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"574","DOI":"10.1016\/0022-2836(94)90032-9","article-title":"Position-based sequence weights","volume":"243","author":"Henikoff","year":"1994","journal-title":"J. Mol. Biol"},{"key":"2023041105225738200_","article-title":"Untersuchungen zu dynamischen neuronalen Netzen","volume-title":"Diploma thesis, Institut f\u00fcr Informatik, Lehrstuhl Prof. Brauer, Tech. Univ. M\u00fcnchen","author":"Hochreiter","year":"1991"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"1735","DOI":"10.1162\/neco.1997.9.8.1735","article-title":"Long short-term memory","volume":"9","author":"Hochreiter","year":"1997","journal-title":"Neural Comput"},{"key":"2023041105225738200_","article-title":"Gradient flow in recurrent nets: the difficulty of learning long-term dependencies","volume-title":"A Field Guide to Dynamical Recurrent Networks","author":"Hochreiter","year":"2001"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"518","DOI":"10.1002\/prot.20221","article-title":"Remote homolog detection using local sequence-structure correlations","volume":"57","author":"Hou","year":"2004","journal-title":"Proteins Struct., Funct. and Bioinformatics"},{"key":"2023041105225738200_","first-page":"149","article-title":"Using the fisher kernel method to detect remote protein homologies","author":"Jaakkola","year":"1999"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"846","DOI":"10.1093\/bioinformatics\/14.10.846","article-title":"Hidden markov models for detecting remote protein homologies","volume":"14","author":"Karplus","year":"1998","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","first-page":"656","article-title":"BLAT \u2013 the BLAST like alignment tool. Genome Research","volume":"12","author":"Kent","year":"2002"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"527","DOI":"10.1142\/S021972000500120X","article-title":"Profile-based string kernels for remote homology detection and motif extraction","volume":"3","author":"Kuang","year":"2005","journal-title":"Journal of Bioinformatics and Computational Biology"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"467","DOI":"10.1093\/bioinformatics\/btg431","article-title":"Mismatch string kernels for discriminative protein classification","volume":"20","author":"Leslie","year":"2004","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"95","DOI":"10.7551\/mitpress\/4057.003.0007","article-title":"Inexact matching string kernels for protein classification","author":"Leslie","year":"2004","journal-title":"Kernel Methods in Computational Biology"},{"key":"2023041105225738200_","first-page":"225","article-title":"Combining pairwise squence similarity support vector machines for remote protein homology detection","author":"Liao","year":"2002"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"2224","DOI":"10.1093\/bioinformatics\/btl376","article-title":"Remote homology detection based on oligomer distances","volume":"22","author":"Lingner","year":"2006","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"4321","DOI":"10.1093\/nar\/gkf544","article-title":"A comparision of profile hidden Markov model procedures for remote homology detection","volume":"30","author":"Madera","year":"2002","journal-title":"Nucleic Acids Res"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"536","DOI":"10.1016\/S0022-2836(05)80134-2","article-title":"SCOP: a structural classification of proteins database for the investigation of sequences and structures","volume":"247","author":"Murzin","year":"1995","journal-title":"J. Mol.Biol"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"1201","DOI":"10.1006\/jmbi.1998.2221","article-title":"Sequence comparisons using multiple sequences detect three times as many remote homologues as pairwise methods","volume":"284","author":"Park","year":"1998","journal-title":"J. Mol. Biol"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"2444","DOI":"10.1073\/pnas.85.8.2444","article-title":"Improved tools for biological sequence comparison","volume":"85","author":"Pearson","year":"1988","journal-title":"Proc. Natl Acad. Sci"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"4239","DOI":"10.1093\/bioinformatics\/bti687","article-title":"Profile based direct kernels for remote homology detection and fold recognition","volume":"21","author":"Rangwala","year":"2005","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"265","DOI":"10.1093\/bib\/3.3.265","article-title":"PROSITE: A documented database using patterns and profiles as motif descriptors","volume":"3","author":"Sigrist","year":"2002","journal-title":"Brief. Bioinform"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"195","DOI":"10.1016\/0022-2836(81)90087-5","article-title":"Identification of common molecular subsequences","volume":"147","author":"Smith","year":"1981","journal-title":"J. Mol. Biol."},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"401","DOI":"10.1093\/bioinformatics\/14.5.401","article-title":"Reduced space hidden Markov model training","volume":"14","author":"Tarnas","year":"1998","journal-title":"Bioinformatics"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"4673","DOI":"10.1093\/nar\/22.22.4673","article-title":"CLUSTAL W: improving the sensivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice","volume":"22","author":"Thompson","year":"1994","journal-title":"Nucleic Acids Res"},{"key":"2023041105225738200_","article-title":"The Nature of Statistical Learning Theory. Statistics for Engineering and Information Science","author":"Vapnik","year":"2000","edition":"2nd edition"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"131","DOI":"10.7551\/mitpress\/4057.003.0009","article-title":"Local alignment kernels for biological sequences","volume-title":"Kernel Methods in Computational Biology","author":"Vert","year":"2004"},{"key":"2023041105225738200_","doi-asserted-by":"crossref","first-page":"513","DOI":"10.1093\/bioinformatics\/btg005","article-title":"Alignment-free sequence comparision\u2013a review","volume":"19","author":"Vinga","year":"2003","journal-title":"Bioinformatics"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/23\/14\/1728\/49813447\/bioinformatics_23_14_1728.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article-pdf\/23\/14\/1728\/49813447\/bioinformatics_23_14_1728.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2024,2,14]],"date-time":"2024-02-14T04:31:08Z","timestamp":1707885068000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/23\/14\/1728\/189356"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2007,5,11]]},"references-count":35,"journal-issue":{"issue":"14","published-print":{"date-parts":[[2007,7,15]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btm247","relation":{},"ISSN":["1367-4811","1367-4803"],"issn-type":[{"value":"1367-4811","type":"electronic"},{"value":"1367-4803","type":"print"}],"subject":[],"published-other":{"date-parts":[[2007,7,15]]},"published":{"date-parts":[[2007,5,11]]}}}