Over the past few years, the annual prevalence of ulcerative colitis (UC) has almost doubled, and for Crohn’s disease (CD) it’s almost tripled. These forms of inflammatory bowel disease (IBD) and others affect millions across the globe, with diagnosis being slow, invasive, and expensive. Plus, the latest scientific evidence shows that in Western societies these conditions tend to affect a younger population now than they have in the past.

The last 15 years have also seen a substantial growth in the fields of machine learning and metagenomics, which made PMI believe that something further can be done to aid IBD patients. In 2019, this led PMI to create the data science sbvIMPROVER Metagenomics Diagnosis for Inflammatory Bowel Disease Challenge (MEDIC), aimed at bringing in the wisdom of crowds and the scientific community to find better diagnostic tools for IBD.

Schematic Description of the Challenge

You’ll be tasked with answering these questions, which could give millions a fast non-invasive diagnosis, leading to a more comfortable life: Are metagenomics data sufficiently informative to predict IBD status? Which predictive computational model is the most accurate to classify patients as IBD vs non-IBD, or CD vs UC? What do the most discriminative metagenomics features tell us?


This year’s collaboration has the potential to represent a step forward in IBD diagnosis and provide healthcare professionals with AI-based tools/technology to enhance the accuracy and speed of the diagnostic process.


Who is the challenge for?

Anyone – whether in a research group, start-up, company, or independent – who wants to help advance microbiomics and patients with IBD is welcome to participate in the sbvIMPROVER challenge.

This year’s challenge is now live and will run until January 15th, 2020. You’ll be given three large metagenomics datasets from the sequenced DNA of stool samples of non-IBD subjects, and patients with and without IBD (including CD and UC).

With this, you’ll be tasked with identifying the most accurate computational methods for distinguishing between IBD and non-IBD patients, and submit the discriminative signatures used in your predictive model.

sbvIMPROVER challenge 2019 timeline

After the challenge closes, submissions from every participating group will be anonymized and scored according to the gold standard labelling of IBD/non-IBD patient groups, then the top teams or individuals will share the $12,000 prize pool.

Display your data science skills, grow your professional network, collaborate with industry/academic peers, and above all help us work towards designing new diagnostic tools for people suffering from IBD.

Click here to learn more and join the challenge.


What is sbvIMPROVER?

sbvIMPROVER is a collaborative project created to provide quality control of industrial research by verifying the relevant methods. The focus on process verification via crowdsourcing makes sbvIMPROVER stand out, compared to other projects that center only on basic scientific questions.

PMI Science servers

Crowdsourcing is group-oriented collaboration where anybody can contribute to an open project. The people involved often have wildly diverse backgrounds, which surfaces methods and solutions that may otherwise be hard to bring under the spotlight. Data science in particular, as a discipline, still in its infancy has greatly benefited from this approach over the past few years in projects ranging from GWAS and cancer research to the CAMI initiative and our challenges from previous years.

For the last decade, AI and metagenomics have been at the forefront of innovation. Combining both of these rapidly evolving fields with large datasets and the most important element of all – you – gives this challenge the opportunity to genuinely make a landmark difference in the lives of IBD patients.


sbvIMPROVER challenge history

Previous sbvIMPROVER challenges have covered a range of topics, collated over 400 participants from 65 countries, and resulted in 13 peer-reviewed publications. We’ve noted past challenges below and picked a few key publications for your interest.


Prior sbvIMPROVER challenges:

  • Microbiomics – Identifying taxonomic profiles (2018)
  • Network verification – Verifying lung biology network models (2018 & 2014)
  • Japan Interpretation datathon – Assessing biological impact on omics data (2017)
  • Israel Epigenomics – Predicting labels from epigenomic/transcriptomic data (2017)
  • Singapore datathon – Analysing big data in epigenomics (2016)
  • Systems toxicology – Differentiating smokers from non-smokers using gene expression data (2015 & 2014)
  • Species translation – Are biological events in rodents translatable to humans? (2013)
  • Diagnostic signature – Computational approaches for disease phenotype prediction based on transcriptomics (2012).


Key publications:

  • Titz B et al., Proteomics and Lipidomics in Inflammatory Bowel Disease Research: From Mechanistic Insights to Biomarker Identification. 19. E2775. Int J Mol Sci. 2018.

  • Scotti E et al., Exploring the microbiome in health and disease: Implications for toxicology. Tox Res and Appl. 1. 2017.

  • Poussin et al., Crowd-sourced verification of computational methods and data in systems toxicology: a case study with a heat-not-burn candidate modified risk tobacco product. Chem Res Toxicol. 30. 934-945. 2017.

  • Belcastro et al., The sbv IMPROVER Systems Toxicology Computational Challenge: Identification of Human and Species-Independent Blood Response Markers as Predictors of Smoking Exposure and Cessation Status. Computational Toxicology. 5. 38-51 2017.

  • Rhrissorrakrai et al., Understanding the limits of animal models as predictors of human biology: lessons learned from the sbv IMPROVER Species Translation Challenge. 31. 471-483. Bioinformatics. 2015.

  • Poussin et al., The species translation challenge - A systems biology perspective on human and rat bronchial epithelial cells. Scientific Data. 2014.

  • Tarca et al., Strengths and limitations of microarray-based phenotype prediction: lessons learned from the IMPROVER Diagnostic Signature Challenge. Bioinformatics29. 2892-2899. 2013.