The Biocomputing Research Lab

Image: Figures pertaining to research

Dose-Response Characterization in Phenotypic Assays


Identification of lead compounds against complex parasites has traditionally involved animal models of infection and in vitro whole-organism, phenotypic assays. Automation of the latter remains a significant challenge hampering progress in hit and lead identification. We present an approach to automatically determine dose-response characteristics for chemical screening using biological imaging and machine learning. This method is made publicly available through the QDREC (pronounced 'QuadRec') web server. QDREC utilizes human experts to train a classifier to distinguish parasites that are affected by drug-action from those that exhibit normal characteristics. The classifier(s) can thereafter be utilized in a completely automated manner. We demonstrate the applicability of our approach in the context of drug screening against one of the etiological agents of schistosomiasis.

We have made the data and software underlying QDREC available under the AGPLv3 open-source software license. Data, code and documentation related to QDREC may be found here.

The available files are listed below:

  • Archive containing a snapshot of the first QDREC release, including the complete project files and attendant documentation.
  • Browesable source files, including documentation and pre-release updates, are hosted on GitHub.
  • Archive of images from the main dataset. The main dataset contains 258 images depicting approximately 11,309 schistosomula, representing four complete sets of concentration points for each drug (plus controls). This dataset was divided into two subsets for training (5,511 parasites, '-a' series images) and testing (5,798 parasites).
  • Spreadsheet listing all segmented parasites from the main dataset, along with their with automatic and manual classifications.
  • Readme documenting parasite spreadsheet files
SF State Home