SBI2 Special Interest Group at SLAS 2018

The Society for Biomolecular Imaging and Informatics and SLAS will co-host the 2018 HCS/HCA Data and Informatics SIG on Wednesday, February 7th, 12PM-1:15PM in the San Diego Convention Center, San Diego, CA.


The format will be a guided discussion of three HCS/HCA Data and Informatics topics led by discussion leaders who will present a brief introduction with some background and data slides on their themes to prompt the audience to both engage and participate in a lively discussion.

Theme: “Concerned about the Analysis of Multiparameter HCS/HCA Data: find out what Deep Learning can do to help!”

Topic 1: "Beyond the conventional information in images"

Discussion Leader: Minh Doan, Ph.D. Imaging Platform, Broad Institute of MIT and Harvard, Cambridge

Modern bioimaging is rapidly changing, including the expansion of dimensionality at both image acquisition and data analytics stages, and the arrival of deep learning that reshapes the feature space. We will present recent efforts to leverage these techniques in a variety of applications.

Topic 2: “Deep Learning Analysis of High Content Imaging Screens”

Discussion Leader: Dana Nojima, Ph.D. Genome Analysis Unit, Amgen, Inc.

High Content Imaging screens produce phenotypically rich data sets. To leverage this complexity, detailed image analysis measuring hundreds of features with subsequent multivariate analysis have been utilized. Recently Deep Learning workflows based on Convolutional Neural Networks (CNN) have demonstrated their usefulness as a tool for analysis of High Content image data.

Topic 3: “Deep Learning for HCS: quick understanding of phenotypic space, reliable classification results and easy analysis transfer”.

Discussion Leaders: Stephan Steigele & Matthias Fassler, Genedata AG, Basel, Switzerland.

We’ll provide a 5-minute introduction on the main steps for applied deep learning in the HCS domain illustrating the associated paradigm shift. We’ll depict its huge potential and together with the audience we’ll challenge three aspects: generation of highly resolved maps of phenotypic space, the reliable generation of pharmacologically relevant results and the transfer of image analysis protocols across different specimens and/or imaging modalities.