BioPhotonics spoke with Christian Olsen, associate vice president and industry principal of biologics at Dotmatics, which produces software connecting science and data. Olsen discussed trends in genomics, the challenges and opportunities with AI and bioinformatics, and their real-world applications in health care. Can you describe how AI is being used in the modern life sciences and biomedical industries? The use of artificial intelligence (AI) has changed over the years, particularly in health care. It used to be applied to passively analyze data, but now it’s taken on an active and predictive role in gene sequence editing and analysis, protein structure prediction, and systems biology. We’re working with top biotech companies, and the work ultimately involves knocking down data silos to aid in the digital transformation process. The data might be unstructured, or we may try to flatten the structure of the data and label it. This helps the organization prepare for using AI/ML [machine learning] and also improves data flow in the research and development process of discovering new therapeutics, developing cell lines, designing experiments, and making better decisions. It is hard to be formulaic in biology, and it is also hard to bring engineering principles into working with organisms and their functioning. But AI can help with predicting protein structures, and the binding locations of proteins. This information on how these structures form can be useful in devising treatment strategies. It is only possible to process so many targets, but predictive analysis can help widen that target field. One specific example of how this can work is in genomics and tumor profiling by tracking the sequence of RNA expressions and the tumor microenvironment. You can start answering the question, “How is the tumor evolving in 3D space?” What are some of the challenges that are faced when a business is considering the use of AI? When designing a solution for a particular environment, there are a number of factors at play. You must consider the confluence of an organization’s business process and the data your research and development organization produces every day. You can’t code around a bad business process, and often an inefficient data infrastructure has been created by patching things together; you must figure out when patching is no longer enough. But you can’t ignore that at the heart of the matter are people, and you must consider the input of different stakeholders. You also have to remember ethical and legal issues, and fair data principles, and that an organization’s IP is usually wrapped up in their data. For us, the goal is to see patterns and make connections you couldn’t make when the data existed in silos. What are some of the ultimate goals of software development in biomedicine? One of the things we’ve been involved with is developing a scientifically aware multimodal data platform that gathers information from proteomics, flow cytometry, and other forms of analyses, which come together to help plan therapeutics. But this platform needs to be functional and cater to how research and development is happening in the lab. The platform will help execute scientific workflows and allow various nonscientific business stakeholders to generate reports; it’s designed to be an organic process. Many of the labs and companies we work with are working on a wide variety of therapeutics, scientific hardware, and data types. When we zoom out, we realize some of these details have significant implications for global health and how diseases are formed and successfully treated.