Bibliomics-based Selection of Analgesics Targets through Google-PageRank-like Algorithm

To screen for effective and non-addictive pain-killers in substitution of morphine, appropriate drug targets should be selected firstly. They can be retrieved through transcriptomic, proteomic, bibliomic and any other high-throughput technique. In the view of bibliomics, the known or potential analgesics targets are hidden in all human genes (proteins) cited in MED-LINE entries with "pain" as the major topic. But a satisfying selection of analgesics targets may not be realized due to "noises" generated in the entry recognition procedure of gene symbols. To mine analgesics targets from the literature contaminated by these "noises", we looked at the abstracts in bulk and simply counted how many times each pair of genes or their products were mentioned together in one publication. So a gene's co-citation network had been constructed with biological meaning confirmed for each of the connections. In this network, we had extended the idea from Google in accomplishment of a PageRank-like algorithm, which ranked the drug target capacity of each gene in bibliome. The algorithm allowed efficient prioritization of genes based on a premise that a gene should be highly ranked if other highly ranked genes link to it. It not only gave a good performance in recall of known drug targets, but also made reasonable prediction of the potential analgesics targets from the bibliome.