biomed data mine spatial structur sequenc dbase featur protein classif retriev spatial model classif corneal shape structur dbase queri protein wavelet abil retriev molecul base structur similar applic diseas diagnosi treatment drug discoveri design paper present method repres protein molecul fast flexibl effici retriev similar structur base global local attribut begin comput pair wise distanc amino acid transform structur distanc matrix normal matrix specif size appli wavelet decomposit gener set approxim coeffici serv global featur vector transform reduc dimension data preserv spatial featur correl test method run queri protein dataset previous literatur base comparison label scop databas find method significantli outperform exist approach term retriev accuraci memori util execut time specif tree run nearest neighbor search dataset protein averag accuraci scop superfamili level total queri time time faster previous publish techniqu addit process queri base global similar propos innov extens effect match protein base sole share local substructur allow flexibl queri interfac structur dbase queri protein wavelet abil retriev molecul base structur similar applic diseas diagnosi treatment drug discoveri design paper present method repres protein molecul fast flexibl effici retriev similar structur base global local attribut begin comput pair wise distanc amino acid transform structur distanc matrix normal matrix specif size appli wavelet decomposit gener set approxim coeffici serv global featur vector transform reduc dimension data preserv spatial featur correl test method run queri protein dataset previous literatur base comparison label scop databas find method significantli outperform exist approach term retriev accuraci memori util execut time specif tree run nearest neighbor search dataset protein averag accuraci scop superfamili level total queri time time faster previous publish techniqu addit process queri base global similar propos innov extens effect match protein base sole share local substructur allow flexibl queri interfac model dbase approach visual classif decis patient diagnosi classif biomed data model dbase spatial averag multi dlevel approach scop fold recognit classif protein base structur plai import role deduct discoveri protein function low number solv protein structur unknown relationship structur sequenc requir altern method represent classif effect larg number potenti fold problem classif strategi increas likelihood classifi reach local optima distinguish structur categori present hierarch strategi structur classif partit protein base scop class attempt assign protein fold dataset deriv popul scop fold sequenc base descriptor properti input featur test number classif method includ na ve bay boost strategi achiev averag fold recognit significantli higher previous report literatur indic effect multi level approach altern represent distanc matric character protein structur approach protein structur mine align largest area bioinformat data mine research protein domain effort includ protein structur predict fold pathwai predict sequenc align ab initio simul structur align substructur detect substructur detect gener defin mine molecul structur order find interest frequent domain sequenc align involv determin similar protein molecul base amino acid sequenc match potenti pitfal solv problem case substructur mine focus sole structur inform lead discoveri biolog irrelev substructur sequenc align align result vari greatli depend substitut matrix paper describ method combin benefit substructur mine sequenc align attempt determin similar protein molecul absenc biolog inform work quickli effici mine protein molecul order determin frequent local structur addit biolog sequenc inform algorithm align protein similar local structur sequenc yield global align molecul present structur mine align algorithm addit work cluster metric amino acid base physico chemic properti metric align algorithm order provid mechan global align protein molecul discov spatial relationship approxim equival pattern contact map 