ProFClass-TM
Protein Functional Classification using Text Data-mining

School of Informatics and Faculty of Life Sciences
University of Manchester

"Protein functional classification using text data-mining" is a BBSRC project (2002-2005) carried out jointly by the School of Informatics and Faculty of Life Sciences, University of Manchester, UK.

[1] Overview

Automatic information extraction and analysis of biological text is becoming increasingly feasible, because of the growth of available information from on-line journals, bibliographic databases, web pages and other sources. The volume of written material pertaining to the function of proteins/genes is increasing rapidly, and - some might argue - the quality of the information is decreasing.

Historically, classification has been at the heart of biology and with the increasing volume of data, automatic methods for classifying biological entities are highly desirable. Placing entities into a classification can be an intensive activity.  Invariably, computers assist a manual classification; examples include using sequence homology to infer the functional class of a protein, or retrieval of relevant documents to aid a human classifier.

ProFClass-TM aims to use automatic text-classification to assist in the assignment of proteins to functional categories. Classifying bodies of text (documents) is an active area of research and has applications in information extraction, information retrieval and information filtering. This project involves the application of techniques from text classification - notably Support Vector Machines (SVMs) - to classify proteins into functional classes based on retrieved text documents in combination with experimental and other data. The aim is to develop tools that can accurately predict/extract information on protein function such as sub-cellular location, enzymatic mechanism, and physiological role from combinations of relevant text, sequence, and experimental data.

Textual information on protein function is assembled from a variety of sources and placed in a database. Using the vector model of information retrieval, we use support vector machines and other methods to classify the proteins into functional categories - training on the MIPS classification, Gene Onotology, and Enzyme Registry. The aim is to generate a tool that allows a user to submit a body of text relevant to a protein and retrieve probable functional classes for that protein.

[2] Objectives

• to devise, implement and evaluate methods for automatic functional classification of proteins from relevant text, in order to accelerate the rate of assimilation of information by scientists and consequently facilitating their research.
• to evaluate which methods of text categorization perform  best. Methods will include nearest neighbours, support vector machines and latent semantic indexing alone or in combinations.  The most desirable term-weighting and kernel functions will be determined.
• to apply the successful methods to genomes, genes and proteins which have yet to have function assigned to them.
• to make the resource available to the research community through a WWW portal.

[3] People

• Dr Ben Stapley, principal investigator, Faculty of Life Sciences
• Prof John Keane, co-investigator, School of Informatics
• Dr Goran Nenadic, lecturer, School of Informatics
• Paul Dobson, research assistant, School of Informatics
• Sarah Dyer, PhD student, Faculty of Life Sciences
• Simon Rice, PhD student, Faculty of Life Sciences

[4] Related publications

• B.J. Stapley, L.A. Kelley, and M.J.E. Sternberg: Predicting the Sub-Cellular Location of Proteins from Text Using Support Vector Machines, Proceedings of PSB 2002 (pdf)
  
  
• Nenadic, G., Rice, S., Spasic, I., Ananiadou, S., Stapley, B.: Selecting Text Features for Gene Name Classification: from Documents to Terms, Proceedings of the ACL 2003 Workshop on NLP in Biomedicine, ACL 2003, Sapporo, Japan, 2003, pp. 121-128 (pdf)
  
  
• Nenadic, G., Spasic, I., Ananiadou, S.: Terminology-Driven Mining of Biomedical Literature, in Bioinformatics Vol. 13, 2003, pp.1-6 (link)
  
  
• Rice, S., Nenadic, G., Stapley, B.: Protein Function Assignment using Term-based Support Vector Machines:BioCreative Task Two 2003, Workshop on Critical Assessment of Information Extraction systems in Biology (BioCreAtiVe), Granada, Spain (slides)
  
  
• Krauthammer, M., Nenadic, G.: Term Identification in the Biomedical Literature, Journal of Biomedical Informatics (Special Issue on Named Entity Recognition in Biomedicine), Vol. 37(6):512-526, 2004 (pre-print)
  
  
• Rice, S., Nenadic, G., Stapley, B.: Mining protein function from text using term-based support vector machines, BMC Bioinformatics 2005, 6(Suppl 1):S22, "Critical assessment of text mining methods in molecular biology" (full text)
  
  

[5] BioCreative 2004

Using the technologies devised within the project, we have taken part in the BioCreative challenge. More information at:

• BioCreative challenge 2004 Workshop
  
  
• A critical assessment of text mining methods in molecular biology: Reports
  
  
• Rice, S., Nenadic, G., Stapley, B.: Mining protein function from text using term-based support vector machines, BMC Bioinformatics 2005, 6(Suppl 1):S22, "Critical assessment of text mining methods in molecular biology" (full text)
  
  
• Rice, S., Nenadic, G., Stapley, B.: Protein Function Assignment using Term-based Support Vector Machines:BioCreative Task Two 2003, Workshop on Critical Assessment of Information Extraction systems in Biology (BioCreAtiVe), Granada, Spain (slides)
  
  

---

Last modified: June 1, 2005

---