Functional Annotation of Genes Using Hierarchical Text Categorization

1
Functional Annotation of Genes Using Hierarchical
Text Categorization

• Svetlana Kiritchenko, Stan Matwin University of
  Ottawa, Canada
• and
• A. Fazel Famili
• National Research Council of Canada

2
Functional Annotation of Genes from Biomedical
Literature
3
Previous Research

• Raychaudhuri et al. (2002)
• BioCreative workshop (2004)
• No hierarchical information has been used

4
Advantages of Hierarchical Approach

• Additional, potentially valuable information
• Relationships between categories
• Flexibility
• High levels general topics
• Low levels more detail
• Hierarchical evaluation
• Give credit to partially correct classification

5
Hierarchical consistency

• if (dj, ci) ? True,
• then (dj, Ancestor(ci)) ? True

c1
c1
c2
c2
c3
c3
c5
c5
c4
c4
c7
c6
c7
c6
consistent
inconsistent
6
Hierarchical Local Approach
c1
c2
c3
c5
c4
c7
c6
c8
c9
7
Hierarchical Local Approach
c1
c2
c3
c5
c4
c7
c6
c8
c9
8
Hierarchical Local Approach
c1
c2
c3
c5
c4
c7
c6
c8
c9
9
Hierarchical Local Approach
c1
c2
c3
c5
c4
c7
c6
c8
c9
10
Hierarchical Local Approach
c1
c2
c3
c5
c4
c7
c6
c8
c9
consistent classification
11
New Global Hierarchical Approach

• Make a dataset consistent with a class hierarchy
• add ancestor category labels
• Apply a regular learning algorithm
• AdaBoost
• Make prediction results consistent with a class
  hierarchy
• for inconsistent labeling make a consistent
  decision based on confidences of all ancestor
  classes

12
New Hierarchical Evaluation Measure

• Precision/Recall considering all ancestors of a
  correct (predicted) category
• Simple, straight-forward to calculate
• Based solely on a given hierarchy (no parameters
  to tune)
• Gives credit to partially correct classification
• Discriminates by distance and depth
• Allows to trade off between classification
  precision and classification depth

13
Results