Guan Lab

Department of Computational Medicine & Bioinformatics

GuanLab wins 2014 DREAM RA Drug Responder challenge

The GuanLab team developed a novel data mining approach for predicting drug response using genetic and clinical information. This algorithm ranks top in the 2014 DREAM (Dialogue for Reverse Engineering Assessments and Methods) rheumatoid arthritis drug response challenge, for both genetics-only model and combined model, for both sub-challenges, and for both the leaderboard and the final previously unseen test set.
A winning algorithm of the 2013 DREAM8 Network Inference Challenge

A winning algorithm of the 2013 DREAM8 Network Inference Challenge

We developed an elegant model for predicting the rewiring of causal networks under unseen interventions. This model achieved the best aggregated performance (in accuracy) in the 2013 HPN-DREAM8 Breast Cancer Network Inference Sub-Challenge 2A and 2B: Time-course prediction. It was identified as 'most consistent performing' for this task, defined as ranked robustly first using random subsets for testing. Additionally, this model is orders of magnitudes faster compared to contemporary methods, realizing genome-wide causal network reconstruction and accurate drug response prediction in less than 30 seconds.
Modeling functional relationships for alternatively spliced isoforms through heterogeneous data integration

Modeling functional relationships for alternatively spliced isoforms through heterogeneous data integration

We developed a novel multiple instance learning-based probabilistic approach that integrates large-scale, heterogeneous genomic datasets including RNA-seq, exon array, protein docking and pseudo-amino acid composition for modeling a global functional relationship network at the isoform level in the mouse, through formulating a gene pair as a set of isoform pairs of potentially different properties. The local networks reveal functional diversity of the isoforms of the same gene.
Differentiating functions at the Isoform level

Differentiating functions at the Isoform level

We developed a generic algorithm that interrogates public RNA-seq data at the transcript level to differentiate functions for alternatively spliced isoforms. For a specific function, our algorithm identifies the ‘responsible’ isoform of a gene and generates classifying models at the isoform level instead of at the gene level. This work was selected for 'Highlights' in ISMB2014.