Dream Challenges

CD2H DREAM Challenges

CD2H DREAM Challenges

CD2H and DREAM have partnered to host a series of DREAM Challenges focused on topics relevant to the CTSA Community including phenotype and outcome prediction, NLP, image classification, and other quantitative areas of biomedicine. Idea Competition Update
 (Image 1 of 4)
 (Image 2 of 4)
 (Image 3 of 4)
 (Image 4 of 4)

What is a DREAM Challenge?

The DREAM Challenges are an open science and collaborative challenge framework. A DREAM Challenge poses one or more questions within a scientific domain to the biomedical research community to spur innovative solutions. DREAM has successfully run over 50 Challenges in multiple disease and biomedical domains spanning basic, translational, and clinical research.

Examples of previous Challenges include:

• Prediction of drug sensitivity in cell lines using genomic data. [link]
• Prediction of patient outcomes in metastatic prostate cancer using clinical trial data. [link]
• Assessment and comparison of variant detection algorithms from DNA sequencing. [link]
• To learn more about previous DREAM Challenges please visit the DREAM Challenge home page

CD2H DREAM Challenges and the CTSA Program

The CD2H is bringing DREAM Challenges to the CTSA Program to help promote the collaborative development and dissemination of innovative informatics solutions to accelerate translational science and improve patient care. The CTSA Program houses a wealth of innovative ideas, data, and clinical and translational expertise. The CD2H DREAM Challenges will help leverage these strengths and provide a platform to facilitate cross hub collaboration and productive industry partnerships. The benchmarks, algorithms, and other solutions developed through the CD2H DREAM Challenges will help advance the use of cutting-edge informatics across the CTSA Program and beyond.

Opportunities for engagement with CD2H DREAM Challenges:

• Propose a scientific question through the CD2H Idea Competition (see below)
• Champion a challenge
• Contribute data
• Form a team and participate in an active DREAM Challenge
• Pilot a DREAM Challenge solution at your hub

CD2H Idea Competition for the CTSA Program

Semi-Finalists Announced

We are very pleased to announce the semi-finalists for the CD2H Idea Competition! Congratulations to the following teams and their proposals:

  • Predicting risk of retinopathy of prematurity (ROP) in premature infants using vital sign data and growth parameters, Columbia University
  • Predicting respiratory illness following preterm birth to guide therapy choices and expectations, University of Rochester
  • Analysis of brain tumor imaging for clinical trial development, Georgetown University
  • Dynamic Stratification of Venous Thromboembolism Risk in Cancer Patients, UC Davis
  • Automated Scoring of Radiographic Damage in Rheumatoid Arthritis, University of Alabama
  • Enabling methods for accessible and reproducible quantitative MRI of the liver, University of Wisconsin-Madison

Thank you to all the teams that submitted! We received many outstanding applications (32 total), and hope to develop more opportunities to support CTSA Challenges in the near future.

We will be following up with each of the semi-finalists, requesting additional information that will need to be incorporated in updated proposals as we move forward with the next round.

Competition Timeline

Next stages of competition:

March 15, 2019
6 finalists announced

April 19, 2019
Revised proposals due

May 3, 2019
Show and Tell virtual presentations from 6 finalists

May 9, 2019
Winning proposal announced

What is the Idea Competition?

The Idea Competition is designed to solicit ideas and proposals from the CTSA community for a future DREAM Challenge. The winner of the CD2H Idea Competition will be awarded a subcontract for up to $100K to support activities associated with the development and execution of a challenge, that include (a) data extraction and curation (b) algorithm development and assessment, and (c) domain expert support.

The ideal “Idea” proposal should pose a question that is solvable, measurable, and of broad interest to the CTSA biomedical community. Example Challenges might include:

• Prediction of 30-day readmission or patient disease risk.
• Inference of patient diagnosis.
• Assessment of Natural Language Processing algorithms against a gold-standard data set.
• Comparison of clinical trial matching algorithms.

RFP Information

Eligibility criteria
• Proposals accepted from any researcher, clinician, or staff member affiliated with a CTSA hub with a current UL1 Award.
• Proposals must include a letter of support from the affiliated CTSA hub’s principal investigator.
Individuals affiliated with sites that receive subcontracts from CD2H are not eligible for this competition.

Application Requirements
• 1-2 page proposal, addressing the questions listed here
• Letter of support from the principal investigator at your CTSA hub
• Submit your proposal and letter of support to data2health@gmail.com with the subject line: Idea Competition Application
• Deadline for initial proposals and letters of support was February 15, 2019, 5 p.m. per local time of the applicant

Proposal Review Criteria

Scientific questions
• Relevance: address a meaningful and high-interest problem in the biomedical community.
• Solvability: poses a question that is answerable given existing data and knowledge.
• Measurability: results in an answer (or prediction) that can be assessed for correctness against a gold-standard.

Data
• Most Challenges are “data challenges” that require data sets that are used for the training and validation of algorithms. Proposals will be evaluated on having identified relevant data sets – in the public and/or private domain – that can be used by challenge participants to develop and assess algorithms.

Ability to carry out the challenge
• Appropriate domain or methodological expertise

Support for the Winner

• The winner of the CD2H Idea Competition will receive a subcontract up to $100k. The subcontract will be awarded directly to the researcher, clinician, or staff member who will be playing a leading role in running the associated DREAM Challenge.
• Subcontract funds to support the involvement of the winning team in running the challenge in collaboration with CD2H. Funds can be used to cover:
• Data acquisition and curation costs
• FTE support

Please send all questions to data2health@gmail.com