Threading using neural nEtwork (TUNE): the measure of protein sequence-structure compatibility. Export

@article{citeulike:230867, abstract = {MOTIVATION: Fold recognition programs align a probe protein sequence onto protein three-dimensional (3D) structure templates. The alignment between the probe sequence and the most suitable template can be used to predict the 3D structure and often biological function of the probe. Here we present a new threading scoring function of protein sequence-structure compatibility. An artificial neural network model is trained to predict compatibility of amino acid side-chains with structural environments. Log-odds scores of predicted probabilities from this model can then be used to construct protein sequence-structure alignments. RESULTS: Our model is tested on discrimination of native and decoy protein 3D structures. With a residue level structural description, its performance is comparable to those of pseudo-energy functions with atom level structural descriptions, better than the two functions with residue level structural descriptions. AVAILABILITY: The C++ source code of our neural network model is available at http://mathbio.nimr.mrc.ac.uk/\~{}kxlin.}, address = {Division of Mathematical Biology, National Institute for Medical Research, The Ridgeway, Mill Hill NW7 1AA, UK.}, author = {Lin, K. and May, A. C. and Taylor, W. R.}, citeulike-article-id = {230867}, citeulike-linkout-0 = {http://view.ncbi.nlm.nih.gov/pubmed/12376379}, citeulike-linkout-1 = {http://www.hubmed.org/display.cgi?uids=12376379}, issn = {1367-4803}, journal = {Bioinformatics}, keywords = {fold\_recognition, neural\_network, threading}, month = {October}, number = {10}, pages = {1350--1357}, posted-at = {2005-11-23 17:34:28}, priority = {2}, title = {Threading using neural nEtwork (TUNE): the measure of protein sequence-structure compatibility.}, url = {http://view.ncbi.nlm.nih.gov/pubmed/12376379}, volume = {18}, year = {2002} }

TY - JOUR ID - citeulike:230867 L3 - citeulike-article-id:230867 N2 - MOTIVATION: Fold recognition programs align a probe protein sequence onto protein three-dimensional (3D) structure templates. The alignment between the probe sequence and the most suitable template can be used to predict the 3D structure and often biological function of the probe. Here we present a new threading scoring function of protein sequence-structure compatibility. An artificial neural network model is trained to predict compatibility of amino acid side-chains with structural environments. Log-odds scores of predicted probabilities from this model can then be used to construct protein sequence-structure alignments. RESULTS: Our model is tested on discrimination of native and decoy protein 3D structures. With a residue level structural description, its performance is comparable to those of pseudo-energy functions with atom level structural descriptions, better than the two functions with residue level structural descriptions. AVAILABILITY: The C++ source code of our neural network model is available at http://mathbio.nimr.mrc.ac.uk/~kxlin. IS - 10 JF - Bioinformatics SN - 1367-4803 AD - Division of Mathematical Biology, National Institute for Medical Research, The Ridgeway, Mill Hill NW7 1AA, UK. EP - 1357 TI - Threading using neural nEtwork (TUNE): the measure of protein sequence-structure compatibility. VL - 18 SP - 1350 KW - fold_recognition KW - neural_network KW - threading AU - Lin, K AU - May, AC AU - Taylor, WR PY - 2002/10// UR - http://view.ncbi.nlm.nih.gov/pubmed/12376379 ER -