Artificial Intelligence is revolutionising various aspects of life sciences, from drug discovery and disease diagnosis to education, showing great promise in improving healthcare and enhancing learning. But, caution is needed in its application, considering challenges like lack of generalisation across patient populations and the risk of over-reliance, as human collaboration and thoughtful use remain essential for its successful integration.
Artificial intelligence (AI) is changing the way we approach science, and the field of life sciences has been no exception. Over the last two decades, researchers have been using machine learning and AI-based approaches in research. But the emergence of new AI-based chatbots — ChatGPT, GPT4, and BioGPT — has not failed to start a new buzz. Researchers and educators are still exploring their way around the system, trying to understand how to effectively leverage these systems to ease pedagogy and build a learning framework around these tools.
Recently, certain areas of life sciences and medicine research have witnessed widespread adoption of AI. These include the use of AI in disease diagnosis, identifying disease-related markers (associated with disease conditions or recovery indicators), analysing medical images, repurposing existing drugs in the market for new applications, facilitating clinical trials, designing medical devices, and even developing new drugs.
AI in drug discovery
Researchers are still testing AI and building algorithms for more efficient and accurate use in biological systems. However, even this development of knowledge and outcomes has been much faster than traditional methods that rely on labour-intensive and expertise-reliant processes. For instance, this year, the first fully AI-generated drug entered clinical trials in human patients i.e., phase 2 clinical trials. The drug is being tested for the treatment of a rare, progressive chronic lung disease called idiopathic pulmonary fibrosis. The drug was designed and clinical-trial-ready in just one year, a record improvement over the current average five-year timeline. AI played a crucial role in designing the drug and discovering the specific protein target that the drug binds to for alleviating the disease.
AI in disease diagnosis
Two other areas of life sciences research and application where AI can have a huge impact are medical image analysis and disease diagnosis.
Phaneendra K. Yalavarthy, Professor, Department of Computational and Data Sciences, Indian Institute of Science (IISc), develops AI based platforms for medical imaging. “AI has revolutionized biomedical imaging by enabling faster and more accurate analysis of medical images such as MRIs, CT scans, and X‑rays,” he says. “It has significantly improved disease detection, diagnosis, and treatment planning, leading to better patient outcomes and reduced medical errors.”
In a country like India, where there is a severe shortage of radiologists to read medical images, AI in medical imaging is a need of the hour, Yalavarthy explains. This would help the radiologists shorten their image-reading times and provide better patient care. “AI is primarily an adjunct at this point of time and evolution of AI in the next decade will push it to driving the patient care,” he adds.
Another group of researchers at IISc combined AI with Raman spectroscopy to quickly detect bacterial pathogens present in different types of clinical samples in a study published last year. One of these groups also worked on using Raman spectroscopy and AI to detect COVID-19, in collaboration with All India Institute of Medical Sciences (AIIMS), Bhopal, for which they won the Challenger Award at the NASSCOM AI Game Changer Awards in 2021.
Raman spectroscopy gives a unique spectrum for each bacterium based on how they scatter light shone on them as a laser beam. The subtle differences in the spectrum arising from the different molecular signatures in different pathogens is difficult to detect with human eye and would also take longer time. But AI helps in rapid and accurate detection of the spectra once trained.
Thus, AI-based tools can improve life and health when integrated in different areas, from research to diagnosis, drug discovery, and treatment. But there are also certain areas that one must treat with caution. For instance, Yalavarthy explains that the lack of generalisation across the patient population is one of the challenges in using AI. He adds that most AI algorithms do not perform consistently across patient populations, and it would take another decade for AI to have universal applicability. “That is why AI is still only a recommending system,” he says.
In my view, healthcare still requires human touch and AI-based technology alone will not provide the impact. The collaboration between AI and expert will be key for healthcare.
AI in life sciences education
Beyond research and diagnosis, AI is making significant strides in transforming life sciences education. Recent and popular developments in AI are the chatbots ChatGPT, GPT4, and BioGPT. These are large language models that are trained on huge amounts of text available across different domains and languages. For example, ChatGPT is trained on all the text available on the internet until 2021. So, when a text is given as a prompt, the tool taps into all the textual data that it has been trained on and returns answers. GPT4 is essentially an updated version of ChatGPT, which can also accept images as prompts. The explosion of these language models has got researchers and educators thinking about their appropriate use, misuse, and the ability to distinguish between human-written and machine-generated content.
Sachin Rajagopalan, Assistant Professor, Department of Microbiology, Ramnarain Ruia Autonomous College, Mumbai, concurs that AI-based tools like ChatGPT have revolutionised education and can readily answer students’ queries and solve their problems. “But you need to have some prior knowledge,” he warns. “While the tools are good at paraphrasing content and enriching its language, one needs to be vigilant of the content.”
According to him, these are “good outlining tools” that students can use to plan “what they need to cover in their assignments, and perhaps generate some project ideas too.” He adds that some teachers also use this feature to prepare lesson plans.
In addition to these language models, there are AI-based tools that have proven to be useful in creating content for pedagogical tools. For instance, educators could use AI-based tools to convert text into PPT or search for images available under the creative commons license. “These are surely helpful, and several teachers are using this feature,” Rajagopalan says.
A word of caution
Although AI-based tools bring such tremendous benefits for students and educators alike, they also have downsides. One must realise that tools like ChatGPT are language models, and use them accordingly. The answers are as good as the prompts one would provide. In the context of its use in academic pursuits, the tool often returns non-existent reading material and references. Moreover, there are ethical considerations to bear in mind. To what extent should students use these tools? When they do, how best can they use it for their growth? These are questions that are plaguing educators and students. The former are now challenged to prepare questions that tools like ChatGPT cannot easily return, and many are now working on using the tool to their benefit.
As Rajagopalan worries, heavy reliance on AI could handicap the unique nature of human thought process, and therefore hamper intellectual progress. The usefulness of AI ultimately rests in the hands of its user. As Oren Etzioni, an American entrepreneur aptly said,
AI is a tool. The choice of about how it gets deployed is ours.