Automated Insight
How AI-based research tools can empower researchers and dive deeper into data
| 4 min read | Interview
Sherry Winter is Director of Biology and Biomedical Solutions at Elsevier, where she oversees product development and works with customers to understand research challenges before developing solutions and tools that will overcome those barriers. Alongside her more than ten years’ experience in life science product development and a PhD in medical genetics from the University of Toronto, Canada, Winter brings expertise from Qiagen and the Moffitt Cancer Center and Research Institute.
Elsevier recently launched a research tool that “visualizes a comprehensive landscape of biological relationships.” Here, Winter delves into exactly how it empowers researchers.
Broadly, how can digitalization improve the drug discovery process?
Digitalization of manual and paper-based processes is a major step towards more efficient drug discovery and is a journey that most life sciences organizations started a while ago. But it’s not enough to just give researchers access to digitalized internal data. To prevent duplicated experiments and make more confident hypotheses, researchers must be equipped with search tools that enable them to sift through external data sources too, such as journal articles, public data repositories, regulatory documents, and clinical trial data.
Why are AI tools so important for the future of research?
Tools with AI capabilities can help relieve the pressure on researchers when it comes to understanding disease biology. Many researchers will read articles to form an understanding, but AI tools can quickly collate critical data from millions of full-text journal articles, abstracts, and thousands of clinical trials. This feat is simply not possible for researchers to do manually. The best research tools are also accessible and structure data in an interactive way. After all, there’s no use bringing datasets together, if the information is too challenging to query or analyze.
What do you mean by a “comprehensive landscape of biological relationships?”
Pharmaceutical companies are setting their sights on solving diseases with increasingly complex biology, where many genes, proteins, tissues, and processes are involved. Yet intense competition demands that insights into these relationships are identified quickly and confidently to bring new therapies to patients sooner. Researchers need to be equipped with access to data and analytical skills to achieve this level of speed.
In any given disease or drug mechanism, there is a vast range of related biology concepts and relationships that a researcher needs to understand to advance their work; for example, gene expression; proteomics/physical interaction; metabolomic/molecular transport and modification; functional association (between a disease and a cellular process or another disease); and regulation.
AI-based research tools highlight the links between these to allow researchers to understand disease biology and make AI-informed decisions based on crucial evidence. This information is displayed in a Sankey diagram that makes relationships easy to visualize.
Upstream connections are displayed to the left (things that impact on the term) and downstream connections to the right (concepts are acted on by the search term). For example, if a researcher searched for a potential drug target, “EPAS 1” (Endothelial PAS domain-containing protein 1, also known as Hif2a), to the left they could identify potential antagonists; for example, proteins upstream of EPAS1 that negatively affect its expression. To the right, they can see diseases that EPAS1 is associated with causing. This format means researchers can understand disease development, progression, and drug responsiveness, whatever their level of data skills.
How else can researchers focus on specific targets in such a huge database?
Tools can (and should) include features that can help. Our tool has filters that allow researchers to narrow down their research questions and confirm their experimental hypotheses for new drug targets, biomarkers, and drug repurposing projects. Researchers can look for new findings by filtering for relationships only reported once or twice – or filter by specific publications dates, specific clinical trial information, or type of relationship (such as proteins that positively or negatively regulate a disease).
How do you expect such research tools to continue to advance in the future?
Life sciences companies are rapidly investing in technology that can help them automate data curation and literature review. Machine learning and other AI-based technologies are at the top of this hype cycle, and are being hailed by many as the future of more efficient drug discovery and research. However, it’s important to recognize that generalized AI-tools aren’t the best fit for life sciences research since they lack subject matter expertise and nuance. If companies really want to accelerate research over the coming years, they should consider using research tools that are purpose-built for drug discovery and the pharmaceutical industry.
