10545

9728

كارشناسي ارشد

هوش مصنوعي

اصفهان: دانشگاه صنعتي اصفهان، دانشكده برق و كامپيوتر

1394

سيزده، 119ص.: مصور

رسول موسوي

چنگيز اصلاح چي

1394/08/03

ناصر قديري، مجيد طالبي

1396/10/05

كتابنامه

برق و كامپيوتر

مهندسي برق و كامپيوتر

ID9728

چكيده از نقطه نظر زيست شناسي كامپلكس ها به عنوان واحدهاي اصلي فرآيندهاي تتند ايم گروه هاي پروتئيني در تي برخوردار هست تتي از اهميت خاصت زيست فرآيندهايي همچون رو نوي سي و تكثير DNA ت سريع واكنش هاي زي ستي ساير فرآيند هاي زي ستي نقش سلولي و سيگنال ها چرخه حيات هدايت ا سا سي دارند ك شف گروه هاي پروتئيني موجود در كنار پيش بيني و ت شخيص گروه هايي كه هنوز شناختهن شدهاند سبب مي شود درك بهتر و كامل تري از ا صول ساختاري و رفتاري سلول ها حا صل شود اگرچه روش هاي آزماي شگاهي شده ا ست اما محدوديت هاي شنا سايي ايم گروه ها م رفي مختلفي براي مختلف و خا صي نيز در ايم بيم موجود ا ست نكته قابل توجه در ايم باب تد كه هر كدام توانايي ت امل با وجود هزاران پروتئيم گوناگون مي باشت تخيص تكل بزرگ پيشرو تشت پروتئيم هاي ديگر را دارند به ايم ترتيب مشت گروه هاي پروتئيني اصلي از ميان ميليون ها گروه پيشنهادي مي باشد به ايم ترتيب روش هاي جايگزيم محاستتباتي را مي توان مكمل هاي مفيد و سودمندي براي روش هاي آزماي شگاهي دان ست طي دهه گذ شته مقدار ب سيار زيادي از داده هاي مربوط به ت امل هاي پروتئيم پروتئيم مربوط به گونه تگاهي قدرتمندي جمع تط تكنيك هاي آزمايشت هاي مختلفي از موجودات توست آوري شده اند ايم داده ها كه م موال به صورت شبكه هاي ت املي پروتئيم پروتئيم نمايش داده مي شوند ايم امكان را مي سر مي سازند تا بتوان كامپلكس ها را از دل آن ها كش تف نمود رويكردي كه در مقابل روش هاي ت آزمايشگاهي مي تواند هزينه بسيار كمتري داشته باشد به ايم دليل كه ايم شبكه ها را مي توان به صورت گرافي مدل كرد كه پروتئيم ها راس هاي آن و هر يال نماينده اي براي ت امل ميان يك جفت پروتئيم باش تد ايم ت توانايي وجود دارد تا مسئله اصلي به مسئله خوشه بندي گراف ها در علوم

Proposing a new protein complexes detection method in PPI networks based on eliminating noise of network via effect of removing hubs Ali Mohammad Ali Maddi ali maddi@ec iut ac ir 2015 Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 IranDegree M Sc Language FarsiSupervisor Dr Sayyed Rasoul Mousavi Assist Prof srm@cc iut ac ir Advisor Dr Changiz Eslahchi Prof Ch Eslahchi@sbu ac ir AbstractBiologically protein complexes are the key molecular entities to perform many essentialbiological functions such as transcription and replication of DNA catalyzing biologicalreactions signal transduction cell cycle and so on Understanding principles of cellularorganization and function can be enhanced if we detect known and predict undiscoveredprotein complexes within the cells While there are a number of ways to detect proteincomplexes experimentally but there are several limitations to this methods The remarkablething is that thousands of different proteins exist which have the ability to interact with otherproteins Thus the major problem is to identify the groups of millions of protein that isproposed Due to these limitations alternative computational approaches for detecting thecomplexes are thus useful complements to the experimental methods for detecting proteincomplexes Over the last decade high throughput experimental techniques have allowed usto collect a large amount of protein protein interaction PPI data for many species Apopular abstraction for representing this data is the PPI network which makes it possible topredict protein complexes from the network Such predictions may be used as an inexpensivetool to direct biological experiments This networks can be represented as undirected graphs in which nodes represent proteins and edges represent interactions between pairs of proteins This networks allow us to tackle the problem of complex prediction with the aid of clusteringtechniques But because of too much noise which include waste edges and incomplete datain such graphs available graph clustering algorithms have not achieved to appropriateresults Also unfortunately many of clustering algorithms have several limitations that arenot suitable for using of PPI networks Such as some of them are designed only forunweighted graphs and some of them assign proteins to the only one group While PPInetworks are modeled in a weighted graph and many evidences have demonstrated that manyproteins belong to more than one main group and protein complexes overlap with each other In this research we propose a novel four part method based on removing hubs to reduce andconsiderable amount of noise in the network This new algorithm utilizes both given edgeweights and can find overlapping clusters So by using of our proposed method complexescan be more accurately distinguished on different data sets than ClusterONE algorithm Keywords Complex detection PPI networks Hubs removal Graph clustering

رسول موسوي

چنگيز اصلاح چي

ناصر قديري، مجيد طالبي