Comparing predictors of disordered protein.

X. Li, Z. Obradovic, C. J. Brown, E. C. Garner, A. Dunker

Research output: Contribution to journalArticle

34 Citations (Scopus)

Abstract

More than 6,000 amino acid sequence attributes were ranked by their conditional probabilities for indicating ordered or disordered protein structure. The top 10 each from several different groups of attributes were merged with still other attributes and then subjected to selection by logistic regression. Evidently, the determination of order or disorder results from the interplay among several attributes, such as average Coordination Number, aromatic content and the numbers of non-polar amino acids, all of which favor the ordered state, and others like Net Charge, Flexibility Index, and the presence of certain polar amino acids, all of which favor disorder. The top 12 selected attributes were used as inputs for artificial neural network (ANN) predictors. Five predictors were developed, compared with each other, and with previous work. The best of these shows substantially improved generalization compared to our previously published predictor.

Original languageEnglish (US)
Pages (from-to)172-184
Number of pages13
JournalGenome informatics. Workshop on Genome Informatics
Volume11
StatePublished - 2000
Externally publishedYes

Fingerprint

Amino Acids
Amino Acid Sequence
Proteins
Logistic Models

Cite this

Li, X., Obradovic, Z., Brown, C. J., Garner, E. C., & Dunker, A. (2000). Comparing predictors of disordered protein. Genome informatics. Workshop on Genome Informatics, 11, 172-184.

Comparing predictors of disordered protein. / Li, X.; Obradovic, Z.; Brown, C. J.; Garner, E. C.; Dunker, A.

In: Genome informatics. Workshop on Genome Informatics, Vol. 11, 2000, p. 172-184.

Research output: Contribution to journalArticle

Li, X, Obradovic, Z, Brown, CJ, Garner, EC & Dunker, A 2000, 'Comparing predictors of disordered protein.', Genome informatics. Workshop on Genome Informatics, vol. 11, pp. 172-184.
Li, X. ; Obradovic, Z. ; Brown, C. J. ; Garner, E. C. ; Dunker, A. / Comparing predictors of disordered protein. In: Genome informatics. Workshop on Genome Informatics. 2000 ; Vol. 11. pp. 172-184.
@article{a716017f4e104ae5b6eebd55b22d6d6f,
title = "Comparing predictors of disordered protein.",
abstract = "More than 6,000 amino acid sequence attributes were ranked by their conditional probabilities for indicating ordered or disordered protein structure. The top 10 each from several different groups of attributes were merged with still other attributes and then subjected to selection by logistic regression. Evidently, the determination of order or disorder results from the interplay among several attributes, such as average Coordination Number, aromatic content and the numbers of non-polar amino acids, all of which favor the ordered state, and others like Net Charge, Flexibility Index, and the presence of certain polar amino acids, all of which favor disorder. The top 12 selected attributes were used as inputs for artificial neural network (ANN) predictors. Five predictors were developed, compared with each other, and with previous work. The best of these shows substantially improved generalization compared to our previously published predictor.",
author = "X. Li and Z. Obradovic and Brown, {C. J.} and Garner, {E. C.} and A. Dunker",
year = "2000",
language = "English (US)",
volume = "11",
pages = "172--184",
journal = "Genome informatics. International Conference on Genome Informatics",
issn = "0919-9454",
publisher = "Universal Academy Press",

}

TY - JOUR

T1 - Comparing predictors of disordered protein.

AU - Li, X.

AU - Obradovic, Z.

AU - Brown, C. J.

AU - Garner, E. C.

AU - Dunker, A.

PY - 2000

Y1 - 2000

N2 - More than 6,000 amino acid sequence attributes were ranked by their conditional probabilities for indicating ordered or disordered protein structure. The top 10 each from several different groups of attributes were merged with still other attributes and then subjected to selection by logistic regression. Evidently, the determination of order or disorder results from the interplay among several attributes, such as average Coordination Number, aromatic content and the numbers of non-polar amino acids, all of which favor the ordered state, and others like Net Charge, Flexibility Index, and the presence of certain polar amino acids, all of which favor disorder. The top 12 selected attributes were used as inputs for artificial neural network (ANN) predictors. Five predictors were developed, compared with each other, and with previous work. The best of these shows substantially improved generalization compared to our previously published predictor.

AB - More than 6,000 amino acid sequence attributes were ranked by their conditional probabilities for indicating ordered or disordered protein structure. The top 10 each from several different groups of attributes were merged with still other attributes and then subjected to selection by logistic regression. Evidently, the determination of order or disorder results from the interplay among several attributes, such as average Coordination Number, aromatic content and the numbers of non-polar amino acids, all of which favor the ordered state, and others like Net Charge, Flexibility Index, and the presence of certain polar amino acids, all of which favor disorder. The top 12 selected attributes were used as inputs for artificial neural network (ANN) predictors. Five predictors were developed, compared with each other, and with previous work. The best of these shows substantially improved generalization compared to our previously published predictor.

UR - http://www.scopus.com/inward/record.url?scp=0034578463&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0034578463&partnerID=8YFLogxK

M3 - Article

VL - 11

SP - 172

EP - 184

JO - Genome informatics. International Conference on Genome Informatics

JF - Genome informatics. International Conference on Genome Informatics

SN - 0919-9454

ER -