This week on Data in Biotech, Ross is joined by Jonathan Eads, VP of Engineering at genomics intelligence company Genomenon, to discuss how his work supports the company’s mission to make genomic evidence actionable.
Jonathan explains his current role leading the teams focused on clinical engineering, curation engineering, platform development and overseeing Genomenon’s data science and AI efforts.
He gives insight into how Genomenon’s software works to scan genomics literature and index genetic variants, providing critical evidence-based guidance for those working across biotech, pharmaceutical, and medical disciplines.
Jonathan outlines the issues with inconsistent genetic data, variant nomenclature and extracting genetic variants from unstructured text, before explaining how human curators are essential to ensure accuracy of output.
Jonathan and Ross discuss the opportunities and limitations that come with using AI and natural language processing (NLP) techniques for genetic variant analysis.
Jonathan lays out the process of developing robust validation datasets and fine-tuning AI models to handle issues like syntax anomalies and outlines the need to balance the short-term need for data quality with the long-term goal of advancing the platform’s AI and automation capabilities.
We hear notable success stories of how Genomenon’s platform is being used to accelerate variant interpretation, disease diagnosis, and precision medicine development.
Finally, Ross gets Jonathan’s take on the future of genomics intelligence, including the potential of end-to-end linkage of information from variants all the way out to patient populations.
Data in Biotech is a fortnightly podcast exploring how companies leverage data innovation in the life sciences.
Chapter Markers
[1:50] Introduction to Jonathan and his academic and career background.
[5:14] What Genomenon’s mission to ‘make genomic evidence actionable’ looks like in practice.
[14:48] The limitations of how scientists and doctors have historically been able to use literature to understand genetic variants.
[16:08] Challenges with nomenclature and indexing and how this impacts on access to information.
[18:11] Extracting genetic variants from scientific publications into a structured, searchable index.
[22:04] Using a combination of software processes and human curation for accurate research outputs.
[24:57] Building high functionality, complex, and accurate software processes to analyze genomic literature.
[29:45] Dealing with the challenges of AI and the role of human curators for the accuracy of genetic variant classification.
[34:37] Managing the trade-off between short-term needs for improved data and long-term goals for automation and AI development.
[38:39] Success stories using the Genomenon platform including making an FDA case and diagnosing rare disease.
[41:55] Predictions for future advancements in literature search for genetic variant analysis.
[43:21] The potential impact of Genomenon’s acquisition of Jack's Clinical Knowledge Base.
