Microsoft powerpoint - text mining - pharma perspective public.ppt

Pfizer Global R&D Sandwich Laboratories Cognitive Bandwidth vs. Domain Complexity • Biology is too big and complex to fit in • We need knowledge about causation• We need knowledge systems as part of – We don’t know what we don’t know– We don’t even know what we do know – To my target?– To my project?– To my disease? – .related to target biology?– .related to secondary pharmacology? • We cannot reliably make discoveries by • Data integration alone is insufficient• We need Systematic context-driven • Biology domain is too big and complex to – Browsing and correlation can’t get us there • We need systems to generate testable – Based on experimental results– Based on literature • We need to use these systems to extract – To build rationale– To identify likely risks Generic drug discovery hypothesis structure: “A drug against Target X will treat Disease Y” “Show me all the diseases associated
with PDE5 from scientific literature”
There’s ~ 6000 curated diseasesHere’s just one: (ASTHMA or asthmatic or Acute severe asthma or Asthmaticus or phosphodiesterase V or phosphodiesterase 12 million
phosphodiesterase-5 or phosphodiesterase PDE(5A) or phosphodiesterase (PDE) 5A or 4 billion words
methylxanthine or zaprinast or tadalafil or vardenafil or SKF-96231 or YC-1 or DMPPO or UK-83405 or Sch-51866 or UK-343664 or WIN-65579 or GF-248 or T-1032 or SR- Statistical co-occurrence and Natural Language Processing used to identify evidence with high PharmaMatrix - US Patent No. US2005060305 17th March 2003 Hopkins el al.
Text-mining and data integration can provide confidence-in-rationale level How do we know we are working on the best projects? Statistical and ontological text-mining can be used to infer new hypotheses – context of the pharma programme is critical • Comprehensive assessment of target and address perceived risks and build rationale – Design experimental strategies to address – Select projects with greater confidence of • Text mining technologies can make an

Source: http://www.nactem.ac.uk/files/phatfile/Ian_Harrow.pdf