Better Than Google™ (if you are a pharmaceutical company) Professor Daniel Miranker Department of Computer Science Institute of Cell and Molecular Biology

Past databases First there were databases …and structured queries “Find all employees near retirement” Then there were search engines… A database cannot query documents. EMPLOYEE NameTitleYears DanProfessor20 …

New opportunity: do both  Current database systems incorporate both for Internet and business settings. Much of the promise of XMLMuch of the promise of XML Documents marked up with data fieldsDocuments marked up with data fields Availability of text field search functionsAvailability of text field search functions  Not so with current biological databases: Data present in many different classes and formsData present in many different classes and forms Explosive growth of data in each classExplosive growth of data in each class Ad hoc search methods for data in each classAd hoc search methods for data in each class Computationally expensiveComputationally expensive

Complex chemical and biological databases Structured properties pose no problems. For many properties, both Google ™ and Oracle ™ are powerless.  3-D chemical fields  Mass spectra Proteins Small molecules

The MoBIoS Solution  Molecular Biological Information System (pronounced Mobius): Organizes data by similarityOrganizes data by similarity Locates matching data in rapid, scalable mannerLocates matching data in rapid, scalable manner Benefit to life science data is analogous to moving text files to a relational databaseBenefit to life science data is analogous to moving text files to a relational database  Relates data from: Protein mass-spectraProtein mass-spectra Small molecule physical propertiesSmall molecule physical properties Combinatorial chemistry librariesCombinatorial chemistry libraries Nucleotide and peptide sequencesNucleotide and peptide sequences Protein structure fieldsProtein structure fields And many other formsAnd many other forms

MoBIoS’ value propositions  Proteomics: Protein identification by mass-spectroscopyProtein identification by mass-spectroscopy  Quantitative protein identification without isotope labeling  Real-time data analysis on desktop computers (vs. batch processing on clusters)  Rational drug design: Real-time query of combinatorial chemistry libraries based on electrostatics and docking modelsReal-time query of combinatorial chemistry libraries based on electrostatics and docking models  New intranet portals: Data-mine existing intellectual property for related compounds and locate internal expertiseData-mine existing intellectual property for related compounds and locate internal expertise

Protein identification by mass-spectrum database search (with E. Marcotte)

Data retrieval by matching electrostatic shape of proteins (with C. Bajaj)

Markets  Market for bioinformatic solutions and hardware: Total: $1.02BB, with growth predicted to $3.0BB in 2010Total: $1.02BB, with growth predicted to $3.0BB in 2010 Software: $0.45BB, with growth predicted to $1.2BB in 2010Software: $0.45BB, with growth predicted to $1.2BB in 2010  Savings from bioinformatic applications: Laboratory and materialsLaboratory and materials  Up to 33% decrease in annual costs for drug discovery efforts  Elimination of expensive laboratory procedures  MoBIoS offers greater functionalities with additional $ savings. TimeTime  Up to 30% reduction in time spent in drug discovery  MoBIos offers further improvements in computational time and broader data access, all at reduced costs. MoBIoS will replace data analysis by large computer clusters with a standard desktop computer.MoBIoS will replace data analysis by large computer clusters with a standard desktop computer. Priobo, Business Communications Company 2005