Releases: Duke-GCB/iMADS
Additional logging fix and dependency upgrades
Fix apache logging in containers
Flask 1.0
New Model File Locations
Changes configuration to sources model files and associated metadata from https://zenodo.org/record/3405516#.XYOYO5NKhhE
See #161
Initial Release
iMADS 1.0.0
iMADS (integrative Modeling and Analysis of Differential Specificity) is a combined computational-experimental strategy to identify and study the differences in DNA-binding specificity between transcription factor (TF) family members, i.e. paralogous TFs.
The iMADS web application and database offers easy access to predictions made using the two types of models implemented in our framework:
- quantitative TF-DNA binding specificity models for individual factors, trained on genomic-context protein-binding microarray (gcPBM) data using a core-stratified support vector regression approach.
- models of differential specificity between paralogous TFs, trained on gcPBM data using weighted least squares regression (WLSR). The WLSR approach allows us to identify genomic sites differentially preferred by paralogous TFs, i.e. sites for which the difference in binding between TFs is larger than the variability observed in replicate experiments.
iMADS functionality is accessed on the web through two modes:
Search Genome
The iMADS database contains genome-wide predictions for human genome versions hg19 and hg38, for eleven TFs and ten TF pairs. These predictions can be visualized in the UCSC Genome Browser, using the track hubs trackhub.genome.duke.edu/gordanlab/tf-dna-binding-predictions/hub.txt and trackhub.genome.duke.edu/gordanlab/tf-dna-preferences/hub.txt, or directly in the iMADS web application, using the graphical interface. Users can explore predictions around transcription start sites (TSSs) of gene, using preloaded or custom gene lists, as well as custom lists of genomic coordinates. When using preloaded gene lists, only the genes/transcripts with predicted binding sites in the selected interval will be shown. When using custom gene lists, all genes/transcripts will be shown, regardless of whether or not they contain predicted binding sites in the selected interval.
Make Predictions
Users can also provide custom sequences as input to obtain predictions according to any binding specificity or differential binding specificity (preference) model in our framework.
All predictions can be downloaded in tab-delimited and csv formats. All data sources used in the iMADS web application are described in the DATA SOURCES tab.