简介:识别蛋白质的细胞的本地化是的潜水艇在基因产品的功能的注解特别地有用。在这研究,我们使用机器学习和探索数据分析(EDA)技术检验并且描绘在九细胞的分隔空间局部性的人的蛋白质的氨基酸序列。代表人的蛋白质的3,749个蛋白质序列的数据集从SWISS-PROT数据库被提取。特征向量被创造捕获特定的氨基酸顺序特征。相对一台支持向量机器,一个多层的视感控器,和一个天真的Bayes分类器,C4.5决定树算法是越过在可靠地预言蛋白质的细胞的本地化基于他们的氨基酸定序的潜水艇的所有九分隔空间的最历久不渝的表演者(平均Precision=0.88;平均Sensitivity=0.86)。而且,EDA图形在每分隔空间描绘了蛋白质的必要特征。作为例子,在血浆膜上局部性的蛋白质有恐水病的氨基酸的更高的比例;细胞质的蛋白质有中立氨基酸的更高的比例;并且mitochondrial蛋白质有中立氨基酸的更高的比例和极的氨基酸的更低的比例。这些数据证明C4.5分类器和EDA工具能为描绘并且预言人的蛋白质的细胞的本地化基于他们的氨基酸定序的潜水艇是有效的。
简介:Biomimetics(orbionics)istheengineeringdisciplinethatconstructsartificialsystemsusingbiologicalprinciples.Theidealfinalresultinbiomimeticsistocreatealivingmachine.Butwhatarethedesirableandnon-desirablepropertiesofbiomimeticproduct?Wherecannaturalprototypesbefound?Howcantechnicalsolutionsbetransferredfromnaturetotechnology?CanweuselivingnaturelikeLEGObricksforconstructionourmachines?Howcanbiologyhelpus?Whatisalivingmachine?Inbiomimeticpracticeonlysome"part"(organ,partoforgan,tissue)oftheobservedwholeorganismisutilized.Apossibletemplateforfuturesuper-organismextensionforbiomimeticmethodsmightbedrawnfromexperimentsinholisticecologicalagriculture(ecologicaldesign,permaculture,ecologicalengineering,etc.).ThenecessarytranslationoftheserulestopracticalactioncanbeachievedwiththeRussianTheoryofInventiveProblemSolving(TRIZ),specificallyadjustedtobiology.Thus,permaculture,reinforcedbyaTRIZconceptualframework,mightprovidethebasisforSuper-OrganismicBionics,whichishypothesizedasnecessaryforeffectiveecologicalengineering.Thishypothesisissupportedbyacasestudy-thedesignofasustainableartificialnaturereserveforwildpollinatorsasalivingmachine.
简介:MicroRNAs(miRNAs)是一个家庭(2123nt)突然,规章的非编码的RNA处理了从长(70110nt)miRNA先锋(pre-miRNAs)。识别真、假的先锋在miRNAs的计算鉴定起一个重要作用。一些数字特征从先锋序列和他们的第二等的结构被提取了适合一些分类方法;然而,他们可以失去在序列和结构隐藏的一些有用地歧视的信息。在这研究,pre-miRNA序列和他们的第二等的结构直接被用来基于在二个序列之间的加权的Levenshtein距离构造一个指数的核。这个字符串内核然后为检测真、假的pre-miRNAs与支持向量机器(SVM)被相结合。在331上基于训练真、假的人的pre-miRNAs的样品,在SVM的2个关键参数被5褶层选择有不同5褶层分区的十字确认和格子搜索,和5条认识被执行。在16独立人士之中,测试从3人,8动物,2工厂,1个病毒,和2人工地假的人设定pre-miRNAs,我们的方法统计上在11个集合上超过以前的基于SVM的技术包括3人,7动物,和1假人的pre-miRNAs。特别地,有通常在以前的工作被排除的多重环的premiRNAs正确地与92.66%的精确性在这研究被识别。
简介:Whencharacterizingflowsinminiaturizedchannels,thedeterminationofthedynamiccontactangleisimportant.Bymeasuringthedynamiccontactangle,theflowpropertiesoftheflowingliquidandtheeffectofmaterialpropertiesontheflowcanbecharacterized.Amachinevisionbasedsystemtomeasurethecontactangleoffrontorrearmenisciofamovingliquidplugisdescribedinthisarticle.Inthisresearch,transparentflowchannelsfabricatedonthermoplasticpolymerandsealedwithanadhesivetapeareused.Thetransparencyofthechannelsenablesimagebasedmonitoringandmeasurementofflowvariables,includingthedynamiccontactangle.Itisshownthatthedynamicanglecanbemeasuredfromaliquidflowinachannelusingtheimagebasedmeasurementsystem.AnimageprocessingalgorithmhasbeendevelopedinaMATLAB?environment.Im-agesaretakenusingaCCDcameraandthechannelsareilluminatedusingacustommaderinglight.Twofittingmethods,acircleandtwoparabolas,areexperimentedandtheresultsarecomparedinthemeasurementofthedynamiccontactangles.
简介:Inthisstudy,wepresentaconstructivealgorithmfortrainingcooperativesupportvectormachineensembles(CSVMEs).CSVMEcombinesensemblearchitecturedesignwithcooperativetrainingforindividualSVMsinensembles.Unlikemostpreviousstudiesontrainingensembles,CSVMEputsemphasisonbothaccuracyandcollaborationamongindividualSVMsinanensemble.AgroupofSVMsselectedonthebasisofrecursiveclassifiereliminationisusedinCSVME,andthenumberoftheindividualSVMsselectedtoconstructCSVMEisdeterminedby10-foldcross-validation.ThiskindofSVMEhasbeentestedontwoovariancancerdatasetspreviouslyobtainedbyproteomicmassspectrometry.BycombiningseveralindividualSVMs,theproposedmethodachievesbetterperformancethantheSVMEofallbaseSVMs.
简介:Nonverbalandnoncontactbehaviorsplayasignificantroleinallowingservicerobotstostructuretheirinteractionswithhumans.Inthispaper,anovelhuman-mimicmechanismofrobot’snavigationalskillswasproposedfordevelopingsociallyacceptableroboticetiquette.Basedonthesociologicalandphysiologicalconcernsofinterpersonalinteractionsinmovement,severalcriteriainnavigationwererepresentedbyconstraintsandincorporatedintoaunifiedprobabilisticcostgridforsafemotionplanningandcontrol,followedbyanemphasisonthepredictionofthehuman’smovementforadjustingtherobot’spre-collisionnavigationalstrategy.Thehumanmotionpredictionutilizesaclustering-basedalgorithmformodelinghumans’indoormotionpatternsaswellasthecombinationofthelong-termandshort-termtendencypredictionthattakesintoaccounttheuncertaintiesofbothvelocityandheadingdirection.Bothsimulationandreal-worldexperimentsverifiedtheeffectivenessandreliabilityofthemethodtoensurehuman’ssafetyandcomfortinnavigation.Astatisticalusertrialsstudywasalsogiventovalidatetheusers’favorableviewsofthehuman-friendlynavigationalbehavior.
简介:TheproblemofflappingmotioncontrolofMicroAirVehicles(MAVs)withflappingwingswasstudiedinthispaper.Basedupontheknowledgeofskeletalandmuscularcomponentsofhummingbird,adynamicmodelforflappingwingwasdeveloped.AcontrolschemeinspiredbyhumanmemoryandlearningconceptwasconstructedforwingmotioncontrolofMAVs.Thesalientfeatureoftheproposedcontrolliesinitscapabilitiestoimprovethecontrolperformancebylearningfromexperienceandobservationonitscurrentandpastbehaviors,withouttheneedforsystemdynamicinformation.Furthermore,theoverallcontrolschemehasafairlysimplestructureanddemandslittleonlinecomputations,makingitattractiveforreal-timeimplementationonMAVs.Boththeoreticalanalysisandcomputersimulationconfirmsitseffectiveness.
简介:Inthepost-genomicbiologyera,thereconstructionofgeneregulatorynetworksfrommicroarraygeneexpressiondataisveryimportanttounderstandtheunderlyingbiologicalsystem,andithasbeenachallengingtaskinbioinformatics.TheBayesiannetworkmodelhasbeenusedinreconstructingthegeneregulatorynetworkforitsadvantages,buthowtodeterminethenetworkstructureandparametersisstillimportanttobeexplored.Thispaperproposesatwo-stagestructurelearningalgorithmwhichintegratesimmuneevolutionalgorithmtobuildaBayesiannetwork.Thenewalgorithmisevaluatedwiththeuseofbothsimulatedandyeastcellcycledata.Theexperimentalresultsindicatethattheproposedalgorithmcanfindmanyoftheknownrealregulatoryrelationshipsfromliteratureandpredicttheothersunknownwithhighvalidityandaccuracy.
简介:蜜蜂(Apismellifera)是有强壮的感觉能力和多样的行为的全部剧目的一只社会昆虫并且为学习学习和记忆的neurobiological基础作为一个好模型有机体被认出。在这研究,我们在microRNA(miRNA)和跟随用下一代的小RNA定序和介绍的Solexa/lllumina数字基因表示标签(DGE)的基于迷宫的视觉学习的送信人RNA(mRNA)分析了变化。为定序的小RNA,我们分别地从迷宫和控制组获得了13367770和13132655个干净标签。40差别的一个总数表示了已知的miRNAs在这二件样品之间被检测,并且所有他们在与控制组相比的迷宫组是起来调整的。为DGE,5681320和5939855个干净标签分别地从迷宫和控制组被检测。有388差别的一个总数表示了在这二件样品之间的基因,与起来调整的45基因和在迷宫组下面调整的343基因,与控制组相比。另外,10差别的表示层次表示了基因被量的反向的抄写聚合酶链反应(qRT-PCR)证实,他们中的八个的表示趋势与DGE结果一致,尽管变化的度在振幅是更低的。miRNA和mRNA表示的综合分析显示出那,在表示的40差别之中已知的miRNAs和388差别表示了基因,miRNA/mRNA的60pairs是在我们的现在的学习共同表示被识别。这些结果建议miRNA和mRNA可以在在蜜蜂学习和记忆的进程起一个枢轴的作用。我们的定序的数据为基于迷宫的视觉学习提供全面miRNA和基因表达式信息,它将便于蜜蜂学习和存储器的分子的机制的理解。