Video Tip of the Week: Leukemia outcome predictions challenge

Although I had other tips in the pipeline, I’m bumping this one up because it is time sensitive. It’s about a competition (or challenge, as they describe it) to use data from cases of leukemia to model make predictions about the outcomes, which could help drive treatment decisions someday. It is called the Acute Myeloid Leukemia Outcome Prediction Challenge.

I found out about it on Google+ via Amina Qutub. In case you can’t see G+, here’s the detail from that post:

Join the crowd to make an impact on cancer research!

The crowd-sourced DREAM 9 Challenge Wiki site opened to all interested scientists, mathematicians, computer scientists, engineers and clinicians around the world. This DREAM Challenge’s goal is to use the wisdom of the crowds to develop new algorithms to understand and treat leukemia, using data provided by M.D. Anderson Cancer Center.

To join, learn about the Challenge, & interact with the data using new visualization tools, visit the DREAM Wiki:!Synapse:syn2455683/wiki/

You can sign up to access the data to begin to work with it. But even before that you can check out the visualization options they provide. This video illustrates a tool they have, which lets you examine specific proteins and specific clinical features, as well as the survival data. (As they note about this tool, though: “The DiBS Data Visualization modules are proprietary, patent-pending tools.”)

From their Data Visualization page, you can click below the video to start looking at that heat map and survival curve data–without having to sign up to access the underlying data. Under the video, click the “BioWheel Interactive Visualization” button to kick the tires a bit.

I started to click around with the visualization tools, I can’t quite figure out what the YWHAE and STK11 heat map patterns mean, they looked very different to me in the visualization.  I have signed up to look at the data itself but I haven’t had a chance to dig any more yet. But it’s available to anyone who agrees to the terms of use, and maybe you can suss out some of the signals that would meet the challenge’s goals:

Subchallenge 1: Determine the best model to predict which AML patients will have Complete Remission or will be Primary Resistant.

Subchallenge 2: For patients who have Complete Remission, predict remission duration.

Subchallenge 3: Predict the overall survival time for each patient

Researchers around he world are collecting lots of data on many disease scenarios. It needs to get closer to patients. Projects like this–with many eyes on it, are a nice way to help us get there–here’s a recent piece about other similar efforts: New platforms aim to obliterate silos of participatory science. There are other challenges from the Sage Bionetworks folks as well. They describe their mission this way:

As a 503c nonprofit organization, Sage Bionetworks’ mission is to catalyze a cultural transition from the traditional single-PI, single-lab, and single-company research paradigm to a model founded on broad precompetitive collaboration. This structure would benefit patients by accelerating development of disease treatments, and society as a whole by reducing the cost of health care and biological research. Sage Bionetworks is actively engaged with academic, industrial, governmental, and philanthropic collaborators in developing this distributed research model.

And there will be more challenges in the future–a reference below explains more of the foundation for these types of efforts. Keep an eye out for them, and hack away.


Boutros P.C., Kyle Ellrott, Thea C Norman, Kristen K Dang, Yin Hu, Michael R Kellen, Christine Suver, J Christopher Bare, Lincoln D Stein & Paul T Spellman & (2014). Global optimization of somatic variant identification in cancer genomes with a global community challenge, Nature Genetics, 46 (4) 318-319. DOI:

Dolgin E. (2014). New platforms aim to obliterate silos of participatory science, Nature Medicine, 20 (6) 565-566. DOI: