Wen Yu

Computational biologist lead providing data ingestion, integration, quality control, assay qualification, statistical analysis and presentation/report for many early phase clinical trials in inflammation and oncology areas. Subject matter expertise in pharmacodynamic and patient stratification biomarkers derived from whole-blood stimulation assay or immunophenotyping by flow-cytometry, expression analysis by microarray, and other assays by CNV, SNP, DNA methylation, Taqman or ddPCR platforms. Biomarkers for drug effect and patient stratification derived from mRNA or protein expression profiling using hierarchical clustering, PCA, GLM, set enrichment and pathway analysis. Aggregation and inquiry of molecular networks or MS/MS spectral library with Hbase and Titan graph database. Interactive filtering and visualization of clinicaltrials.gov+drugbank.ca or IHC images with Microsoft Pivot.

Builder of a company-wide qualitative platform to provide automated proteomic analysis starting from raw MS/MS data. It employed multiple search engines under rigorous statistical consideration leading to very good sensitivity and robustness for peptide and protein ID. The platform also serves as the foundation for quantitative proteomics. Vetting of drug target on whether it’s a cell-surface membrane or soluble protein; or by its preferred expression among tissues or disease states using the experimental evidences compiled from internal and public proteomic resources. Novel application for MS/MS spectral library to proteomics including total clustering before identification, probabilistic assessment of the similarity score, comparison of related but not equal spectra to detect modified peptides or to construct spectral network, and more recently library indexing for interactive lookup of MS/MS. Simulation of MRM protein assay with the aid of spectral library to select the best peptides and fragments and to assess potential interferences in the (RT, m/z) vicinity during in silico design of MRM assay. Responsible for the evaluation and acquisition of ArrayStudio/Server to support high-content biomarker development. Developed departmental requirements toward an experiment annotation and data sharing system. Decision tree analysis of SELDI signature to refute its utility as clinical diagnostic marker. Multiple regression and neural nets modeling and prediction of chromatographic retention time of peptides based on their chemical and physical properties. Genetic algorithm to optimize peptide sequencing from MS/MS spectra.

Responsible for the data analysis of protein and mRNA expression of S. Diversa throughout different growth stages, and the responses of D. Vulgaris to oxidative stress and its correlation with mRNA profiling, operon prediction, metabolic pathway and other bioinformatics tasks. Developer of a novel search engine that was 27% more sensitive for peptide ID than the commercial engines at 1/3 of the error rate, and algorithms for quantitative protein profiling to detect the significant differences from the LC-MS runs prepare for the targeted MS/MS and assemble the list of protein differences. It solved the problems in dealing with complex peptide data, irregular chemical background and drifting LC retention times.

Developer of an integrated pipeline for high-throughput peptide and protein identification (10,000 spectra / day) with novel algorithms for the pre-processing of LC/MS/MS data to maximize information content, matching the MS/MS spectra to putative peptide sequences with objective probabilistic scoring model, and clustering of the equivalent MS/MS spectra to conserve valuable computing and human resources. Informatics leader extensively characterizing cancer and normal cell lines / body fluids for the drug target discovery projects.

Extensive profiling of glycopeptides from recombinant proteins employing differential glycosidase digestion followed by MALDI MS. Developed the micro affinity technique to examine glycoprotein Factor IX using magnetic particles and MALDI-TOF MS. Identified an unknown protein by the combination of mass profiling and de novo sequencing using post-source decay reTOF-MS. Novel scoring model for protein identification were formulated and implemented in a C++ program. Created a web-based "MS Lab" application in Javascript to simulate the mass spectrometry characterization of heterogeneously modified glycoprotein with disulfide bonds. Performed SIMION modeling and hardware modification to significantly improve the efficiency of MS/MS interface of a tandem sector mass spectrometer.