How I use AI to turn failed drugs into new medicines

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript. Ignota Labs, a company based in Cambridge, UK, uses artificial-intelligence technology to determine why drugs have failed in clinical trials, before re-engineering the most promising therapies to give them another shot at making it to the patients who need them. The company was set up in 2021 by Layla Hosseini-Gerami, Jordan Lane and Sam Windsor. In February 2025, they closed a US$6.9-million deal to fund the company, and they have since built a strong pipeline of promising drugs, including treatments for autoimmune diseases and blood cancers. Hosseini-Gerami, the company’s chief data-science officer, has a background in chemistry and bioinformatics, which she combines with data science to understand how drugs affect the body. In April 2025, she was included in Forbes magazine’s ‘30 under 30’ list for European science and health care for her work using AI to accelerate the process of bringing safe drugs to market. She describes how the company, and her role in it, came about. When did you become interested in AI? There were no AI modules on offer when I studied for my undergraduate degree in chemistry at the University of Leeds, UK, between 2014 and 2018. At that time, AI was being used across many industries, but was still an emerging technology. My first exposure to AI came in 2016, during a one-year industrial placement as a machine-learning intern at Optibrium, a drug-discovery software company based in Cambridge, UK. At Optibrium, I was building models to predict the molecular properties of different drugs. Specifically, I focused on pKa — a measure of the acidity of a molecule — which influences a compound’s solubility, permeability and ability to bind to its target. Building these models and then seeing them incorporated into software used by pharmaceutical-industry professionals motivated me to stay in the field and pursue a PhD. Towards the end of my Optibrium internship, I reached out to Andreas Bender, a molecular informatician who has since moved to Khalifa University in Abu Dhabi, and he became my PhD supervisor at the University of Cambridge. I had been looking at different research groups, but he drew me in because he talked about rigour and developing AI that truly has an impact on drug discovery in the pharmaceutical industry. In drug discovery, many compounds are approved on the basis of clinical efficacy, without researchers having a clear understanding of their mechanism of action — for example, the proteins they target or the pathways they regulate. I developed computational algorithms that bridge this gap, identifying the biological targets and pathways that drive a drug’s therapeutic effect. At that time — around ten years ago — it really felt as if we were one of the pioneering groups at the crossover of AI and science, because AI drug discovery hadn’t attained the level of interest it currently attracts. Since then, there has been rapid progress. The ongoing challenge is how to use AI to answer questions about biology and how drugs work in the body, which can be random and unpredictable. How did Ignota Labs come about? At the end of my PhD, I got a LinkedIn message from Lane, who I didn’t know at the time. He had previously been the principal scientist at BenevolentAI, a technology company that uses AI for drug discovery, and had spent a decade working across pharmaceutical and AI biotechnology companies and clinical-research organizations. In the message, he said he wanted to start a company focused on drug safety and needed someone to lead on the interface between chemistry, biology and AI. Windsor, our chief executive, had consulted on projects with Merck and Google DeepMind’s AlphaFold team, as well as projects relating to digital transformation in the NHS, before collaborating with Lane to set up Ignota Labs. Lane and Windsor had come across Bender’s research group and had read my papers. At the time, I was writing up my PhD thesis on using biological and chemical information to improve our understanding of drug mechanisms of action, which subsequently won a 2022 outstanding thesis award from the chemistry department at the University of Cambridge. Their collective experience gave them good insight into the inefficiencies of the field and how AI could be used to solve these. Millions of pounds are spent on developing drugs that can, ultimately, have a less than 10% chance of succeeding. This process also consumes energy and water, and relies on animal research. A stand-out issue for Lane was the failure of drug candidates owing to safety concerns, which can be hard to predict before animal or first-in-human studies. At this point, many drugs are simply abandoned. He felt that my research was best applied to addressing this problem. Ignota Labs was formed at around the time that the biotech boom of 2020–21 crashed. This was caused by rising interest rates in the aftermath of the COVID-19 pandemic and a general withdrawal of venture-capital funding, particularly the exit of investor tourists — those that haven’t conventionally invested in an area — from the life-sciences sector. These factors made fundraising much tougher for early-stage biotech companies and led to higher investor expectations. We needed to refine our strategy and demonstrate both strong science and a clear commercial path. Our audience might not know much about AI or drugs, so we needed to explain complex concepts clearly to investors. We employed communications advisers to help us with this, and we have improved how we convey what we do. We can now also point to our pipeline, to our collaborations with corporations and to case studies that provide evidence that we are able to rescue drugs. How does Ignota Labs do what it does? We have built an AI-driven platform to identify failed drugs and fix their toxicity issues. First, it narrows down thousands of failed drugs to the most promising: the drugs that have huge potential to make a difference to patients but which have hit an unexpected safety issue; for example, causing toxicity in the liver. Next, we apply deep learning, combining bioinformatics, chemoinformatics and multimodal data to understand the root cause of the safety issue and to develop mitigation strategies. Our core aim is to take these drugs, make minimal chemical changes and get them back into the clinic as efficiently as possible. This is different from repurposing, in which the same molecule is tested for a different disease, with no new intellectual-property creation and no opportunity to ‘fix’ any aspect of the drug. Examples of what our technology can detect and fix are off-target binding effects, in which drugs bind to unintended molecules; pharmacokinetics issues, such as the levels of drug elimination from the body; and distribution problems, like a drug getting into the brain when it shouldn’t. The data we work with are very challenging in their complexity. They range from information about molecular structure and biochemistry — such as how strongly a drug and a protein bind to each other — all the way up to data about how a living being responds to a drug. The complexity comes from trying to join all of this up and make sense of it, which is where AI comes in. As such, we aren’t considering different kinds of data each day; rather, we are trying to combine the data to gain insight into how we can rescue drugs. My role is that of chief data-science officer, which I chose in 2022 when we launched the company. I lead the research and development of AI tools that further our mission of turning drugs around. This includes sourcing data sets to feed our algorithms and overseeing the development of our SAFEPATH platform. Enjoying our latest content? Log in or create an account to continue Access the most recent journalism from Nature's award-winning team Explore the latest features & opinion covering groundbreaking research