The Rise of AI in Biology: A Scientist’s Perspective on What’s Next
About Author: Prof. Debarka Sengupta is an Institute Chair Professor in the departments of Computational Biology and Computer Science at IIIT-Delhi and currently serves as the Head of the Infosys Centre for Artificial Intelligence at the same institute. Additionally, he holds the position of Adjunct Professor at Queensland University of Technology-Brisbane (QUT) and the founder of The Sengupta Laboratory. He was awarded the Biological Science Winner under the Merck Young Scientist Award 2023.
Biology today bears little resemblance to what it looked like just a decade ago. The explosion of data, especially from high-throughput sequencing, multi-omics studies, and real-time diagnostics, has not only transformed the kinds of questions we ask but also the tools we turn to for answers. And at the forefront of this shift is artificial intelligence.
What began as an experiment, training algorithms to detect patterns in genomic data, has now evolved into a full-blown shift in how we decode biological systems. Whether it’s predicting protein structures, designing new drugs, or understanding the language hidden within DNA, AI is making biology faster, deeper, and in many ways, more accessible.
Foundation Models: Learning the Language of Life
Foundation models have emerged as one of the most powerful tools in computational biology. These models are trained on large, unlabelled biological sequence datasets, allowing them to uncover the “hidden information” within DNA and protein sequences. By absorbing evolutionary patterns at scale, they help us understand the underlying grammar that shaped genomes and genes.
Models like ESM2 – Evolutionary Scale Model (developed by Meta AI using 15-billion parameters and transformer architecture), Oliveira, and ProtBERT are helping scientists not only to understand proteins but design novel ones. Their ability to learn from raw sequence data and apply that understanding across different biological problems makes them one of the most promising AI tools in this space.
AlphaFold and Beyond: Cracking the Protein-Folding Code
No discussion about AI in biology is complete without mentioning AlphaFold. Developed by DeepMind, AlphaFold made a significant breakthrough by predicting 3D protein structures from sequence data with remarkable accuracy. Its performance in the CASP 2020 (Critical Assessment of Structure Prediction) challenge marked a turning point in protein structure prediction, an area that had remained largely unsolved for decades.
Its performance in the CASP 2020 (Critical Assessment of Structure Prediction) challenge was a watershed moment. For the first time, computational predictions rivalled experimental results in accuracy, compressing what once took years of wet-lab work into hours. AlphaFold’s impact is profound: it has enabled thousands of previously unsolved protein structures to be modelled, opening up new frontiers in understanding disease mechanisms and designing therapeutics.
Since then, we’ve seen the emergence of AlphaFold-Multimer and AlphaFold3, expanding the capability to protein complexes and protein-DNA interactions. In parallel, open-source alternatives like RoseTTAFold and OpenFold have democratized access to these breakthroughs, catalyzing innovation across academia and industry alike.
“AI is helping us identify drug targets, screen compounds, and even repurpose existing drugs.”
AI in Drug Discovery: Generative Design and Rapid Screening
Drug discovery is another area being transformed by AI, often behind the scenes. What used to take four to five years can now be condensed into a single year, thanks to AI-driven molecular design. DSP-1181, a compound for treating obsessive-compulsive disorder, became the first AI-designed drug to enter clinical trials, all within 12 months.
AI is helping us identify drug targets, screen compounds, and even repurpose existing drugs. Baricitinib, originally developed for rheumatoid arthritis, was repurposed for COVID-19 treatment with help from AI-based predictions. Additionally, drugs like REC-1245 for cancer are being validated in the lab after first being designed in silico. These success stories show that AI is not just generating chemical ideas in silico, but also helping to test ideas far faster than before. Models can now predict which molecules are likely to be effective and safe, focusing researchers’ attention on the most promising candidates and weeding out dead ends earlier in the pipeline.
Uniting Genomes, Proteomes and Clinical Data: The Power of Multi-Modal AI
Biological research inherently involves diverse data types, from molecular sequences and protein structures to imaging and patient health records. AI’s ability to integrate and analyze such multi-modal data sets has unveiled complex disease mechanisms and improved treatment strategies. For example, combining genomic variants with histopathology images and clinical parameters has enhanced prognosis accuracy in cancer. The growth of large-scale biobanks, such as the UK Biobank, supports these comprehensive analyses.
Current research is moving toward foundation models trained on multi-omics, spatial transcriptomics, and clinical data to characterize cell states and tissue organization at unprecedented resolution. These advances promise to accelerate drug target discovery and optimize therapeutic responses. When variant data is combined with clinical history and tissue images, for example, cancer prognosis becomes far more accurate. Researchers are training multi-modal models using extensive biobanks, such as the UK Biobank, which integrates various types of information, including genomics and spatial transcriptomics.
We’re moving toward a future where AI can map not just the genes or the cells, but the full context of how diseases emerge and how treatments can be optimized. It’s like upgrading from a microscope to a planetary telescope; you see more, and you understand better.
“Integrating AI deeply into biological research without considering the possible pitfalls in AI-derived decisions may lead to incorrect conclusions.”
The Road Ahead: Opportunities and Ethical Considerations
The rise of artificial intelligence in biology points toward a future of extraordinary possibilities. Subsequently, an AI-driven “digital twin” model of cells or whole organisms can be used in simulations otherwise hindered by technical and ethical issues. Such models can revolutionise the scale at which biology is being studied, focusing on nano-timescales and infinite what-if scenarios. AI can also guide and accelerate synthetic biology, designing new enzymes or organisms for enhanced functional activities. Integrating genetics and lifestyle (collected through numerous health gadgets), AI can boost preventive measures for non-communicable and lifestyle-related diseases by providing optimal dietary interventions.
As said earlier, with great power comes great responsibility. Integrating AI deeply into biological research without considering the possible pitfalls in AI-derived decisions may lead to incorrect conclusions. Such “drawbacks” raise critical ethical and scientific considerations towards the full implementation of AI as a researcher. Typical AI models are called “black-box” models as interpretability remains a key concern. Other key issues in fully integrating AI include a lack of effective validation screens (except in drug discovery) and hallucinations (false positive associations). Blindly trusting an AI’s output could lead to errors or even harm, so transparency and robust evaluation are paramount.
Other key concerns arise from the lack of suitable data used for developing AI models. Data bias is typically identified in large biobanks (data limited to a few ethnic groups), and when trained on such data, AI-based conclusions might not generalize and could exacerbate health disparities. But with this promise comes certain responsibility.
Additionally, while AI can favourably aid in improving research and life, in the wrong hands, its misuse can lead to biosecurity risks. Novel pathogens and toxins can be used to target communities and countries, forging a base for biological war. Given these ethical concerns, numerous experts were engaged in AI-based teams to review potential risks.
AI has moved from being a tool to becoming a collaborator in biological research. From protein folding to precision medicine, from virtual cells to real-world therapies, AI is helping us answer age-old questions and ask entirely new ones. The next era of biology won’t be defined by isolated discoveries but by interdisciplinary convergence: where biology, computation, and ethics meet. If we guide it well, AI will not just accelerate discovery but fundamentally reframe how we understand life itself.
*The views expressed by the author are his own.
The post The Rise of AI in Biology: A Scientist’s Perspective on What’s Next appeared first on BioVoiceNews.
News