Protein Secondary Structure Prediction by Machine Learning Methods

HROZA, Jiří. Protein Secondary Structure Prediction by Machine Learning Methods. In 1st International Summer School on Computational Biology. Brno, Czech Republic: Masaryk University, 2005, p. 38-43, 5 pp. ISBN 80-210-3907-8.

Basic information

Original name

Protein Secondary Structure Prediction by Machine Learning Methods

Name in Czech

Rozpoznavani sekundarni struktury proteinu metodami strojoveho uceni

Authors

HROZA, Jiří (203 Czech Republic, guarantor)

Edition

Brno, Czech Republic, 1st International Summer School on Computational Biology, p. 38-43, 5 pp. 2005

Publisher

Masaryk University

---

Other information

Language

English

Type of outcome

Stať ve sborníku

Field of Study

10201 Computer sciences, information science, bioinformatics

Country of publisher

Czech Republic

Confidentiality degree

není předmětem státního či obchodního tajemství

RIV identification code

RIV/00216224:14330/05:00014719

Organization unit

Faculty of Informatics

ISBN

80-210-3907-8

Keywords in English

machine learning; protein; protein secondary structure prediction

Tags

machine learning, PROTEIN, protein secondary structure prediction

---

Abstract

V originále

This paper concerns about an application of machine learning methods to a prediction of a secondary structure of an unknown protein. The aim of this study is to the compare artificial neural networks as the state of art method with decision trees and naive Bayes classifier. Detailed experiments are done on selected PDB database data. Results shows that decision trees achieving 87.4 % Q3 accuracy outperform neural networks (80.5 %). Naive Bayes classifier is unusable for this task.

In Czech

Clanek se zabyva aplikaci metod strojoveho uceni na problem predikce sekundarni struktury neznamych proteinu. Cilem je porovnat umele neuronove site, jako nejmodernejsi pouzivane metody, s rozhodovacimi stromy a naivnim Bayesovskym klasifikatorem. Podrobne experimenty jsou provadeny na vybranych datech z PDB databaze proteinu. Vysledky ukazuji, ze rozhodovaci stromy dosahuji mnohem lepsich vysledku (87.4%) nez neuronove site (80.5%). Oproti tomu naivni Bayesuv klasifikator se ukazal jako neprilis vhodny.

---

Links

MSM 143300003, plan (intention)

Name: Interakce člověka s počítačem, dialogové systémy a asistivní technologie Investor: Ministry of Education, Youth and Sports of the CR, Human-computer interaction, dialog systems and assistive technologies