More than 1,000
Reproducible machine-learning experiments conducted
Prioritised candidates
The most promising and differential biomarkers shortlisted
Focused validation
Fewer candidates required further laboratory testing
The Client
Developing a diagnostic test for ovarian cancer
GenoME Diagnostics is a Belfast-based biomarker discovery company developing OvaME, a ddPCR-based diagnostic test for ovarian cancer.
The company had identified a large number of promising RNA biomarker candidates. Before progressing its assay towards commercial use, the team needed to reduce these to a more manageable number for further laboratory validation.
"We had a large number of promising novel biomarkers that we needed to shortlist to a more manageable number, which we planned to take forward for further lab validation and lock-down of the final assay for commercial use."
Shannon Beattie, COO, GenoME Diagnostics
The Decision
Which candidate biomarkers should be taken forward?
GenoME needed a scientifically reasoned and reproducible way to compare its candidate biomarkers and identify which were sufficiently promising and differential to justify further testing.
This required the team to assess its available data, account for potential quality issues and bias, and apply machine-learning approaches across different datasets.
The objective was not to claim that the biomarkers had already been validated. It was to prioritise the candidates that warranted further laboratory validation.
Sonrai's Role
A reproducible approach to candidate prioritisation
Sonrai worked with GenoME to structure and support the analytical process. The work included:
- Assessing the available data sources for quality issues and potential bias
- Creating a framework through which different data types could be brought together
- Designing tailored machine-learning pipelines to investigate diagnostic biomarker candidates
- Applying appropriate privacy measures and access controls
- Providing training and ongoing technical support
- Comparing the results with known biomarkers
More than 1,000 reproducible machine-learning experiments were conducted through Sonrai Discovery, enabling alternative analytical approaches to be run and compared systematically.
The governed environment supported the reproducibility of the work, while Sonrai's team helped GenoME understand the available platform options and use the analytical approach effectively.
"The Sonrai Analytics team were exceptionally helpful. Initially, we did not have a great amount of experience in big data analysis, and the Sonrai team helped to walk us through the platform options and training in a way that was simple to understand, yet professional."
Dr Paul Mullan, Founder, GenoME Diagnostics
The Outcome
A focused shortlist for further laboratory testing
The analysis helped GenoME scientifically, logically and reproducibly shortlist its candidate biomarkers.
Rather than progressing every potential biomarker into hands-on laboratory work, the team could focus its subsequent validation on the candidates considered most promising and differential.
According to GenoME, this accelerated its validation timeline and reduced the hands-on time and cost associated with product development.
What the work established
The engagement produced:
- A structured assessment of the available data
- Reproducible comparison of machine-learning approaches
- More than 1,000 machine-learning experiments
- A reasoned shortlist of potential RNA biomarkers
- A more focused set of candidates for further laboratory validation
The work supported biomarker prioritisation. Final validation of the selected candidates remained part of GenoME's subsequent laboratory programme.
"The results of our engagement with Sonrai has accelerated our validation timeline. It has enabled us to scientifically, logically and reproducibly shortlist our novel potential biomarkers, to only take forward the most promising and differential for further validation in the lab. This has vastly reduced the hands-on time and cost for GenoME Diagnostics product development."
Shannon Beattie, COO, GenoME Diagnostics
Powered by Sonrai Discovery
Reproducible analysis without building an internal data-science function
Sonrai Discovery provided the environment through which the machine-learning experiments could be conducted reproducibly and the analytical results reviewed.
Sonrai combined this infrastructure with training and technical support, helping GenoME apply machine-learning analysis without first assembling an additional internal data-science team.
Which biomarker candidates should your team investigate next?
Sonrai helps diagnostic and drug-development teams evaluate complex molecular evidence, compare candidate biomarkers and determine what warrants further investigation. Speak to our team about the decision your programme is facing.
Discuss your programme