Home

Cytomining Hackathon 2016 - Event theme/summary

For this first-ever cytomining hackathon, 28 attendees (from 11 countries across 4 continents) gathered at the Broad Institute of Harvard and MIT in Cambridge, MA, May 10-12, 2016. The hackathon’s purpose was to bring together one representative from each of the major laboratories working in morphological profiling - in the end, all but 2 targeted laboratories participated.

In morphological profiling, large amounts of quantitative morphological data are extracted from microscopy images of cells to generate a profile; the goal is to identify biologically relevant similarities and differences among samples based on these profiles. Profiles of cell populations treated with different experimental perturbations can be compared to predict a small molecule’s mechanism of action, identify new compounds or materials with desired phenotypic effects, connect unannotated genes to known pathways, infer the functional impact of human genome sequence variants, and identify signatures of disease.

We planned this hackathon because morphological profiling is a wild frontier with novel methods proposed but not yet compared, and certainly no best practices established. Multiple laboratories worldwide are working in this area; this is clearly a time when collaboration would be fruitful. The BD2K/LINCS program is particularly interested in how to process and incorporate morphological profiling information into their data repositories and analysis tools for cross-experiment/comparison. There is thus urgency to bring this community together.

The event met all of our goals - please see the outcome section for details.

Attendance

The morphological profiling community is broad and developing, and we are grateful to have hacked alongside a diverse group of passionate profilers. Participants from India, Colombia, Canada in addition to representatives from all over Europe and the United States attended and shared their profiling methods. Spanning four continents and eleven countries, 28 postdoctoral associates, computational biologists, and software engineers met at the Broad Institute for two days. Twenty-six of these hackers attended in person, while two joined in via videoconference.

Projects resulted / outcomes

1. Knowledge about best practices was shared among nearly every laboratory working in this field.
2. Connections were made among the computational scientists working directly in this field.
3. A living Github wiki was created, with each participant outlining the profiling analysis workflow followed by their own group.
4. The group converged on a plan to publish a paper, “Data analysis strategies for image-based cell profiling” (now published http://dx.doi.org/10.1038/nmeth.4397), which will be an indispensable resource for all of us in the field, describing alternative strategies and best practices for each of the ~20 steps in a typical workflow. Steps accomplished:
   1. Title, strategy/scope, and outline for the paper
   2. Assignment of authors to various sections
   3. Timeline/strategy for completing the paper
5. The group agreed on a test data set and ground truth, for testing methods against each other.
6. The journal Science heard about our hackathon and wrote a news article (http://science.sciencemag.org/content/352/6288/877) covering the development of this field, on “how image-based profiling is being developed as tool for drug, toxin and gene expression function discovery.”
7. Source code was contributed by participant groups for various stages of profiling analysis.
8. The group developed plans for contributing code to existing open source projects.
9. Plans for future meetings were made, starting with summer 2017 (the next meet was in London)

Feedback

Members of the community found it enormously helpful to learn, at the hackathon, how others handle particular challenges in the workflow. Organizing the group around the manuscript outline and pseudocode for the typical workflow was very helpful in clarifying different approaches that can be taken at each step. The group clearly wants to make this an annual event, confirming our suspicion that this would be fruitful. BD2K funds thus jump-started this field and this community.