Panel Discussion Reference Databases Nathan Edwards Georgetown University

Panel Discussion: Reference Databases Nathan Edwards Georgetown University Medical Center

Role of Reference Databases l Provide “exemplar” for comparison w/ metadata: l l l Comparison turns measurements into identity. . . l l Analytical signatures (spectra, fingerprints) Biological knowledge (DNA & protein seq. , SNPs, . . . ). . . match measurements to exemplars (0, 1, many) Identity (shades of grey!) depends on: l l l the nature of the analytical measurement, the model for comparison, and the contents of the reference database. 2

Reference Database Use Cases l Assay design l l l Assay detection l l l Specific detection in device/kit (antibody, microarray) Signal is species specific Non-specific design (mass spectrometry, sequencing) Detection by match to a reference database Assay validation l l l In silico specificity guarantee? Guided specificity testing? In silico validation after update? 3

Assay Design l Specificity: requires all sequences, but l l l Homology and phylogeny l l l mislabeled or incorrect sequences may compromise the design the set of reference genomes will never be complete or "uniformly" sampled may be able to compensate for missing sequences can inform clinical or in silico testing for specificity May lead to overly conservative designs… 4

Assay Detection l l Non-specific design generates sequence or signal from "any" organism Specificity by match to reference database l l Signal and/or reference may not be sequence! May require instrument error modeling May require statistical or error tolerant matches Competition for signal generation? l Testing with mixed (abundance) samples is crucial – not a specificity issue, per se 5

Updated Assay Validation l Specific assays: l l l Target sequence may change (influenza) Updated reference may invalidate working designs Non-specific design assays: l l l Authentication of detection reference Match algorithm changes may affect performance How is instrument/processing software validated across versions? 6