Statistical Bioinformatics Group

Software

Bioinformatics Software and Public Sever System Development Lead:



A.   UVA GEOSS (UVA Internal) & Public GEOSS (Gene Expression Open Source System):
   

• Phase 0 (March 2002) UVA central database server set-up 
• Phase I (Oct. 2002; GeneX 1.5.9): Secured multi-user system construction and validation
• Phase II (August 2003; GeneX Va): (Skeleton of) Tree-structured analysis-interface
• Phase III (Dec. 2003): tight security system with sessioning; automated public posting utility under MIAME guideline
• Phase IV (Dec. 2004; GEOSS, Gene Expression Open Source System): integration of many advanced genomic and computational bioinformatics tools in AnalysisTree interface

 
B. COXEN (COeXpression ExtrapolatioN): Genomic-based personalized chemotheraeutics prediction web system

• Phase 0 (May 2007) UVA central database server set-up (biostat.virginia.edu)
• Phase I (Sep. 2007; COXEN 1.0.2): Public multi-user analysis system for COXEN analysis

C.  Bioconductor packages contributed:

Our group has developed the following software programs for analysis of microarray data.  The programs may be freely distributed but not sold for profit. Also, they are available at the Bioconductor website.  

1. LPE (Local Pooled Error) package. September 2003, with Nitin Jain and Michael O’Connell. For the discovery of differentially expressed genes between two comparing conditions of microarray experiment, performs LPE estimation (based on error pooling/shrinkage methods), test, and rank-invariant resampling-based FDR evaluation (cf. SAM, t-test)
2. HEM (Heterogeneous Error Model) package. May 2004, with HyungJun Cho. For the discovery of differentially expressed genes among multiple comparing conditions of microarray experiment Performs the HEM estimation (using MCMC calculation interfaced with C functions) and provide resampling-based FDR evaluation of H-scores (cf. ANOVA, SAM).
3. MiPP (Misclassification Penalized Posterior) package. March 2005, with Mat Soukup and HyungJun Cho. For the modeling and identification of the most parsimonious classification model on microarray data, performs double-validation discovery (n-fold model training and external cross-validation of model evaluation) of prediction models.