Matching high-throughput genomics to high-throughput mining of the literature

In this study (alas, for-fee-access) , Dennis Wall demonstrates how to mine the literature (aka the biomedical bibliome) to focus the analyses of noisy genomic modalities which by virtue of measuring thousands of genes have to be aggressively corrected for multiple hypothesis testing [ed. Disclosure: I am a co-author]. By examining gene expression analyses of individuals with autism through the lens of the prior literature on neuro-psychiatric-behavioral disorders, he is able to identify genes significantly differentially expressed in individuals with autism, both known and previously non-implicated genes. This is one of a growing list of publications that are attempting to match the high throughput qualities of genomic measurements with an equally efficient automated "reading" of all the painstakingly obtained biomedical investigational literature. It also suggests that an even more detailed annotation by librarians of the existing literature (analogously to what the National Library of Medicine has done for years for the broad addition of meta data) will be productively leveraged in future investigations. I suppose this is where my colleagues from the Semantic Web have another opportunity to feed the search engines of Google.