An integrative methodology for the classification of genes into pathways using a novel text mining approach

DUSSAUT, JULIETA SOL; CECCHINI, ROCÍO LUJÁN; PONZONI, IGNACIO; MAGUITMAN, ANA GABRIELA

Santiago de Chile

Conferencia; ISCB Latin America 2012 Conference on Bioinformatics; 2012

International Society for Computational Biology

This work presents a text mining approach for the classification of genes into pathways. As a starting point the approach uses KEGG´s Pathway database in order to associate genes with its pathways. This data is used for validation and normalization purposes only. The Biocreative II corpus is used to train the prediction algorithm. The first step is to use ABNER to identify biological entity mentions in each of the publications. These entities are used to create a gene frequency matrix. The resulting matrix is then used to compute a co-occurrence weight for each pair of genes. The prediction is then made based on a voting scheme. In order to validate the proposed method we contrast the resulting predictions with KEGG´s Pathway data. As validation metrics we use recall at rank k, precision at rank k and average precision.