An innovative machine learning approach spearheaded by SEBI-Livestock and partners could help advance research on antimicrobial resistance (AMR).
Louise Donnison (senior data scientist) and Isla MacVicar (researcher) recently shared their experiences in using machine learning to accelerate the development of systematic evidence maps with a group of researchers working on antimicrobial resistance. They were speaking at the Symposium on AMR in livestock production in a OneHealth context on 24 May 2023. The Symposium was co-hosted by OECD’s Co-operative Research Programme: Sustainable Agricultural and Food Systems (CRP) and the Edinburgh Antimicrobial Resistance (AMR) Forum, and held at Edinburgh Climate Change Institute.
Louise and Isla introduced the litXpress tool, which emerged from a major effort by SEBI-Livestock to pull together a systematic evidence map for Ethiopia on cattle and small ruminant health and mortality, covering over 700 research articles published over the last decade. Systematic review is a necessary but time-consuming step before conducting research. Reviews often require manually searching, screening and extracting data from PDF documents. A researcher can spend hundreds of hours on one review.
In 2019, SEBI-L reached out to The Bayes Centre (School of Informatics, University of Edinburgh) to explore whether automation could be applied to developing systematic maps. Together they secured funding from the Edinburgh and South East Scotland City Region Deal’s Data-Driven Innovation initiative to develop a tool that uses machine learning to accelerate the this process.
Developing the litXpress approach
Now SEBI-L is working with EDINA, the University of Edinburgh’s centre for data and digital expertise to further develop the litXpress tool and test it in other scholarly domains. Andrew Horne, EDINA project manager, believes it has potential to be scaled up as a general research tool that can help increase research output for the The University of Edinburgh.
During the AMR Symposium, Andrew explained that automation can shave hundreds of hours off the research process. Depending on the review, 50%-90% of the work involved can be automated using the litXpress machine learning approach. Reducing this barrier means more, better research will be initiated, and researchers can spend more time on actual research.
Machine Learning applications to AMR research
During the session, participants discussed how Machine Learning could help accelerate AMR research by mining publicly available databases on a number of subjects:
- AMR-related interventions and their effectiveness (especially in low-and middle-income countries)
- Extracting data from antibiotic sales, import and export documents
- Antimicrobial effectiveness in different animals and diseases
- Linking GenBank database entries with their metadata.
There are many potential applications for machine learning in the AMR One Health field and the litXpress team will continue to explore these opportunities.
Thanks again to OECD’s Co-operative Research Programme: Sustainable Agricultural and Food Systems (CRP) and the Edinburgh Antimicrobial Resistance (AMR) Forum for inviting SEBI-Livestock, and to Carys Redman-White and Prof Dominic Moran for organising an excellent event!
Learn more