Can you share your journey in the field of AI?

 After spending approximately three years as a software engineer, I pursued a PhD program at the Indian Statistical Institute (ISI), focusing on machine learning/data mining and its application in biological problems. ISI is home to some of the country's leading pattern recognition experts, and my experience there was incredibly enriching. At that time, deep learning had not yet emerged, and our focus was on studying clustering and classification techniques as part of classification machine learning/pattern recognition. Since then, I have become engrossed in exploring new applications of machine learning and artificial intelligence across various basic and translational biology domains, with a particular emphasis on cancer. This journey has allowed me to witness the evolution from classical pattern recognition to the development of large language models such as ChatGPT.

What does Single-cell genomics entail? Could you elaborate on your research involving big data algorithms in Single-cell genomics?

 Genomics traditionally involves profiling DNA, RNA, or proteins at the bulk tissue level. Single-cell omics techniques enable the examination of the abundance of various molecules at a single-cell resolution. This provides a more in-depth understanding of biological systems, aiding in appreciating intra- and intratumoral heterogeneity in cancer patients. Given that single-cell data sets can be massive, our laboratory pioneered the integration of big data algorithms to reduce analysis time significantly. We have developed numerous methods to cluster single-cell data, facilitating the identification of rare cells.

What notable advancements does AI contribute to cancer care?

AI has revolutionized the field of medical science, particularly in cancer care. Its applications are widespread, ranging from cancer diagnosis to treatment decision-making. Deep learning techniques have been applied to various scans such as mammography, pathological images, MRI, CT scan, and PET scan surpassing human accuracy. Our group has developed AI applications to recognize multiple types of circulating tumour cells in patient blood by identifying their gene expression patterns. We have also successfully applied AI in personalized therapeutic decision-making, showcasing its effectiveness in laboratory settings and with actual patients.

Can you elaborate on your research focused on early cancer detection?

Our research involved the creation of a tumour-educated platelet gene panel designed for the early detection of cancer in patient blood. This cost-effective test can be conducted in any laboratory equipped with qPCR facilities. Its accuracy underwent validation across multiple patient samples. In collaboration with a prominent U.S. biotech firm, we also established a method for detecting and characterizing single circulating tumour cells in patient blood. Both projects extensively utilized AI. In the first instance, AI was employed to construct a model for recognizing cancer-based on the expression patterns of 11 genes. In the latter, we developed a method for semi-supervised detection of circulating tumour cells, allowing us to identify triple-negative breast cancer cells in patient blood—a challenging task and the first of its kind.

How has the introduction of AI-enhanced the accessibility and quality of healthcare in rural India?

Realizing AI's full potential represents nothing less than an industrial revolution. With its extensive mobile network, India ensures that AI innovations are within reach of almost the entire population. The impact of AI extends beyond urban centres to rural areas, benefiting people across the spectrum. From GPS-based maps to personalized shopping recommendations on e-commerce platforms, conversational agents handling customer care calls, and the application of vast language models in education, weather forecasting, virtual medical assistance, and more, AI is making significant inroads into improving rural healthcare access and quality.

As an educator and researcher, what guidance would you offer aspiring AI enthusiasts in our country?

I recommend that AI enthusiasts prioritize theoretical learning before diving into practical applications to unlock the technology's full potential. Additionally, fostering collaboration with domain experts is essential to fully comprehend the problem statement before starting model development. AI is not a magic solution; it is crucial to understand the problem statement and its practical challenges. It's important to note that AI models developed for academic exercises may not seamlessly translate to real-world applications. Acquiring knowledge about best practices for unbiased evaluation of AI models is also critical. Lastly, as part of the Infosys Centre for Artificial Intelligence, we organize various educational workshops and seminars—stay informed by following our social media handles.