Transcript Document
ReSurfX effect: Dealing individually and combining Current (without ReSurfX) real accuracy <<<< Five data sources at predicted 95% accuracy each - combined ReSurfX Suite of Solutions Adaptive Hypersurface Technology Data source agnostic Novel solution for highly accurate utilization of multi-parametric data Associated solutions that adapts to unpredictable error properties E.g., Normalization, adaptive error correction Data Analyses AND Technology Design Superior Accuracy and Outcome prediction Repeated application enterprise product with a scalable delivery model Cloud Product in Market on Customer Request (customized solutions – currently specific to one tech) Product at: https://vysen.resurfx.com Information at: http://resurfx.com/vysen Affymetrix GeneChip® (a technology for measuring gene expression in large scale) Dominant Market Solutions Too many errors ReSurfX Simplified workflow – Affymetrix™ GeneChip® data Download Visualize Upload from your repository/ Local machine File input Select included GeneChip® ChipType Search/Fetch from public repository NCBI/GEO 1. 2. Fetch CEL files Un-compress Data Analysis Extract 1. 2. 3. CEL file format conversion Extract to usable format Normalization Compare ReSurfX output to a proprietary (Rosetta Resolver™) solution Rosetta 1840 ReSurfX 1081/946 In red: 2 fold cut 948 /983 892 /857 189 /89 Publication 1, 2008 An empirical check of the differentials using an independent approach indicates that most Rosetta specific genes are false positives. These are either what is estimated to be below the sensitivity of the technology with this number of replicates by ReSurfX, or errored due to other limitations. Found by Rosetta but not by ReSurfX 220 7 ReSurfX 529 213 Fold change vs. p-value of genes considered error by ReSurfX 316 Rosetta Publication 2, 2006 Compare ReSurfX output to a widely used public domain (BioConductor & GenePattern) tools Next slide: Note the number of parameters (genes) found by above pipeline but considered error by ReSurfX ReSurfX only > 3 fold (30 of 285) Data of Publication 3 - UP Ctrl Treat 207 37 94 Data of Publication 2 - UP UP Upregulated BioC only > 3 fold (99 of 360) fold change cut problem too Publication 3 - UP Ctrl Treat Publication 2 - UP Publication 3, 2012 (used 3 fold cut) Note the amount of errors now if one uses analysis of Publication 2 and compare results of Publication 3 (cumulative error in a longitudinal pipeline) Unpublished customer data, 2013 (powerful way of easily identifying targets without confounding errors) Among other things: Ctrl 1 Ctrl 2 Treatment 1 Treatment 2 1. RSX found (BioC missed) an interesting kinase, that was genetically checked and proved to be associated with difference in phenotype between Treatment 1 and Treatment 2 2. Other RSX specific changes corroborated as gene families, or gene duplicates BioC & RSX found the kinase against controls Two different studies : Same treatment but slightly different controls (High reprodicibility of ReSurfX analysis) Study 1: 178 146 of 178 overlap (2 fold change cut) overlap increases if no fold cut Study 2: 412 Analysis by ReSurfX Shouldn’t you be using ReSurfX in your data analysis for better outcomes? (save time and money – avoid late stage failures)