Youngsahng Suh explains how AI is advancing the field of clinical diagnostics
The Covid-19 pandemic has underscored the need for speed in the development of tests, vaccines and therapeutics like never before. Behind the scenes, many companies have achieved this with a sophisticated secret weapon: artificial intelligence (AI). Proprietary algorithms have driven everything from faster discovery, optimisation and validation through to more efficient manufacturing, quality control and supply chain management. This is true of vaccines and of clinical diagnostics, which have played a critical role from the earliest days of the pandemic.
How do AI and machine learning (ML) tools help us develop diagnostics quicker? Here, we explore AI’s pivotal role in designing complex assays and advancing them through the validation process. To do this, the platforms maximise today’s rich healthcare datasets and cloud-based computing platforms, reimaging how, what and when diseases can be detected and how individual patients will fare.
Rapid Development and Optimisation
AI and machine learning systems are integral to the development of many modern diagnostics. Although SARS-CoV-2 testing has been front-and-centre during the pandemic, there are many valuable applications in other infectious diseases, cancers and chronic illnesses.
Beyond speed, such applications can drive greater sensitivity and specificity with sophisticated feedback loops. They also help companies adapt as testing needs change – for example, through the emergence of new variants. For example, using automated AI systems, the team at Seegene is able to develop a sensitive and specific molecular diagnostic test in a matter of weeks, instead of years. This is not the result of advances in a single step, but rather the introduction of specialised AI/ML at many points in the development journey.
Seegene often begins with automated referencing software that conducts data mining and pulls references relevant to a new test from all corners of the world. Natural language processing (NLP) technology is used to gather preliminary information, which is then reviewed by the company’s researchers to understand where, medically and scientifically speaking, the diagnostic platform can help the most and where knowledge gaps exist. With this information in hand, the team uses its in-silico lab – powered by an AI-based big data analytics platform – to develop the firm’s own diagnostic tests.
For SARS-CoV-2, an AI-based automated development system can help design and optimise the reagents in the PCR kit, streamlining the test-development process. This can ultimately reduce the work of testing labs to just a few hours, instead of weeks. Machine learning-based platforms can subsequently assist with assay validation before the test is brought to market. Every company has different technology. For Seegene, this step involves its team simultaneously running through approximately 16 processes to ensure that the product works in a real-world setting. These tests, which took just a couple of weeks to perform, would previously have taken months to validate manually.
The Rich Potential of Complex Datasets
Beyond streamlining and optimising the work of researchers and clinicians, AI/ML promises to unlock entirely new applications. With complex datasets and customised algorithms, companies are now entering entirely new terrain – with confidence.
In SARS-CoV-2, multiplex testing has been an important development. Multiplex tests can be used to check for more than one SARS-CoV-2 variant and/or influenza. At an individual level, this allows clinicians to quickly establish what is driving a person’s symptoms and what medical interventions are needed. From a public health perspective, multiplex tests help groups efficiently monitor different infectious diseases and the incidence of specific variants.
Although useful, developing a multiplex test has historically been challenging. Designing, optimising and validating a multiplex diagnostic assay manually would normally take upwards of one year. With AI, that process can be completed in weeks. This creates a lot more flexibility, as companies can rapidly respond to the emergence of new variants with tests that look for those key mutations.
For biotech companies depending on AI for the development of diagnostics, proper data management is key. The management of data is what ultimately determines the success of AI. The Institute of Seegene Information Science is constantly building its competence for data architecture, data governance and data management.
Untapped Biomarker Discovery
Creating diagnostic tests ultimately depends on how fast a biotech firm can handle and process a vast amount of data that is related to virus patterns, illnesses and related treatments. That is largely done with the help of ML and AI. One of the areas where AI/ML looks particularly promising is in the discovery of novel biomarkers for disease.
Biomarkers are tell-tale signs of a biological process that has been dysregulated, helping guide a more definitive diagnosis. But again, discovering clinically relevant biomarkers can take years. By providing computers with the right information, AI can be used to create algorithms that quickly allow companies to discover new biomarkers of disease. Other companies are using algorithms for risk stratification, looking at a series of biomarkers to predict whether or not a person’s cancer is likely to return and if they would benefit from a given therapy.
For chronic illnesses, early detection of relevant biomarkers could prompt lifestyle changes or medical interventions that delay or curb the onset of the disease. With the increased speed, personalisation, and complexity of modern AI-driven diagnostic tests come increased quality of life and health.
Advancing in Leaps and Bounds
Seegene believes that ML and AI will definitely reshape the world we live in. Because we are able to develop diagnostic tests faster with heightened accuracy, we will be able to live in an era where molecular diagnostics become part of our lives. This technology will improve – not replace – the work of lab personnel and clinicians, streamlining their work and unlocking new potential in poorly understood diseases. When that happens, we will be able to prevent not just the pandemic but all kinds of diseases. Molecular diagnostics will work as a precursor in treating and preventing the diseases from happening.
Youngsahng Suh is head of Diagnostic Data Research Center at Seegene