Text-based experiment retrieval in genomic databases

Extraction of biological information from genomic data is very important in the field of bioinformatics. Genomic data is stored in different data formats such as experimental measurements, sequence data and network. With the rapid growth of these data structures in databases such as GEO (Gene Expression Omnibus), ArrayExpress and GenBank, reaching accurate and relevant data has become an important issue. In thisstudy, it is aimed to develop a text-based experiment retrieval framework for the retrieval of relevant experiments from genomic databases. In the proposed framework, Jaccard similarity metric was used as a lexical similarity method, Latent Semantic Analysis (LSA), Probabilistic Latent Semantic Analysis (PLSA) and Latent Dirichlet Allocation (LDA) methods were used as semantic similarity methods. To the best of our knowledge, this is the first study to use semantic similarity methods in the retrieval of time-series microarray experiments. The system performance was tested with the textual information of 120 different time series experiments of Arabidopsis Thaliana plant obtained from the GEO database. With the biological and statistical significance tests, obtained results were verified. As a result, it was observed that semantic similarity approaches were more successful than text-based and content-based retrieval methods in identifying relevant experiments. It is expected that the developed system will shed light on future studies by achieving a successful performance in retrieving relevant experiments. Moreover, the framework can be easily adapted to different types of genomic data.