Semantic Similarity over the Gene Ontology

1
Semantic Similarity over the Gene Ontology
CIKM 2005
Family Correlation and Selecting Disjunctive
Ancestors http//xldb.fc.ul.pt/rebil/

• F. M. Couto, M. J. Silva, P. M. Coutinho

2
Family Correlation

• Family similarity is also a structural similarity
• Like sequence similarity, but
• Overcomes some of the limitations of sequence
  similarity
• Since it is based on experimental results about
  protein domains
• Validate GO semantic similarity with family
  similarity
• Pfam similarity
• a database of protein families
• assigned to UniProt proteins

3
GraSM (Graph-based Similarity Measure)

• Motivation
• Semantic similarity measures use Gene Ontology as
  a hierarchical tree
• Approach
• Selecting multiple ancestors of two GO terms
• Describe different interpretations
• Results

4
Disjunctive Ancestors
(t1,t2) ? DisjAnc(t5) (t0,t1) ? DisjAnc(t5)
Since (t1,t2) ? DisjAnc(t5) then CommonDisjAnc(t4
,t5) t2,t1
5
Cellular Component
Molecular Function
Biological Process
6
Cellular Component
Molecular Function
Biological Process
7
Conclusions

• Family correlation provided a measures ranking
  that is
• uniform over all the different aspects of GO
• consistent with previous studies using different
  corpora
• We have demonstrated the higher effectiveness of
  GraSM for calculating semantic similarities
  between GO terms
• Higher correlation using disjunctive common
  ancestors than only using the most informative
  common ancestor
• All the measures mentioned in this document were
  implemented by FuSSiMeG (Functional Semantic
  Similarity Measure between Gene-Products)
• FuSSiMeG is available on the Web at
  http//xldb.fc.ul.pt/rebil/tools/ssm/

8
(No Transcript)