
Can you describe why and where you studied chemistry, and your subsequent early career?
I always found chemistry fascinating, since my teens – micrograms of a substance can cure you, or they can kill you, it depends on the compound. That made chemistry really powerful in my mind, and I wanted to embark further into this direction, and learn more about it.
I did my diploma in Frankfurt, Germany, in 2002 in the bioinformatics area (predicting signalling peptides), and my PhD in Cambridge, UK, in the cheminformatics area in 2005, on molecular similarity. Both were excellent choices (not that I knew beforehand though, you often only know afterwards in life!), and I am still very grateful to my supervisors, Gisbert Schneider who was then in Frankfurt (I was in fact his first student there), and Bobby Glen in Cambridge, who taught me lots about ‘real-world drug discovery’.
For my postdoc I joined Novartis in Cambridge/MA, on a Presidential Postdoctoral Fellowship – this meant the ability to work on academic research, but with all resources available in a pharmaceutical company, which was another great experience. It also shaped my thinking about the practical relevance of research today, what should be applicable in practice, to improve drug discovery. Jeremy Jenkins, Meir Glick and John Davies were my mentors there, who provided a fantastic environment to learn more, about the analysis of high-throughput screening data, as well as high-content imaging data which became very popular around my time there, in 2006/2007. I can really recommend to keep one’s mind open about academia, big pharma, start-ups (and beyond) – they all have their advantages, and you learn something different everywhere.
Where are you currently working and what is your current position?
I currently have multiple roles, both at the University of Cambridge as a Professor and group leader, but in parallel I was working recently for AstraZeneca in Cambridge (in the computational safety area), Nuvisan in Berlin (setting up their Research Informatics group), and recently we started Terra Lumina, a start-up for using AI in natural product-based drug discovery. So from this it is already quite clear that my strength is also my weakness – I am interested in analysing chemical and biological data in all its breadth, in very different contexts, with respect to its ability enable theselection of compounds that are safe and efficacious in vivo.
What do you like best about your work?
What really motivates me is to use data for decision making in drug discovery – which data gives you a signal related to in vivo efficacy, and in vivo safety? So basically this means ‘putting things into practice’ – understanding which data is good for which purpose, and this is by no means clear from the onset. Often data is generated due to a ‘technology push’ (‘we can sequence, so we should’), instead of letting data generation be guided by the question. Also putting things into practice means to do the step from only aiming for a publication, towards something that is really useful in drug discovery (both of which are very different things!). I cannot claim to always have achieved it perfectly, but I try to get as close to that goal as I can.
Another aspect I find truly motivating in our area of research is its diversity, this can mean diversity from the scientific side (different data types, algorithms, and endpoints), but equally people from different countries, with very different backgrounds. I lived in multiple countries myself as well (at last count I realized I moved 23 times until that stage!), and this is what I always recommend to my students: in science it’s easy to see the world, so go to a new place on the planet, to different science, in your postdoc (or generally next stage in life).
Which of your papers are you most proud of and why?
There are indeed some papers that I quite like and which I am proud of – but for reasons that are maybe not so apparent from the outside. There is a story behind every paper – how the idea was developed, the people you have met and how you met them, and so on.
The first paper I am quite proud of is from my postdoc times (https://www.nature.com/articles/nchembio.2007.53), on the integration of high-content screening data with target prediction. The reason why I quite like that is that the idea really came about during a postdoc seminar at Novartis, where both Daniel Young (the first author of the work) and myself were postdocs at the time. We met at this meeting where Daniel presented his work from the high-content screening side, I thought ‘hey, let’s add target prediction to this for the mode of action side’, and this is what led to a really nice Nature Chemical Biology publication, which combines a (then) relatively new readout type with a mechanistic angle from computational tools.
The other publication is much more recent, while I was a PI in Cambridge, on using small molecules for the targeted differentiation of stem cells into a desired cell type, in this case cardiomyocytes (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4979408/). This work was done by Dr Yasaman KalantarMotamedi as a first author, and it was really fascinating that the computer, using data and algorithms, was able to select small molecules that led to a beating cardiomyocyte under the microscope – so in this way we basically generated a small part of the working heart, using the power of algorithms, chemistry, and biology. I still find this study fascinating when I think about it again.
Apart from that science is really often very incremental – you find one answer, and you have five new questions, so I am not sure this really led to me being ‘proud’ of what I did in particular. In other words, “The more you know – the more you know what you don’t know!” I think to keep the curiosity alive, like when you were young, or in your PhD, this is what keeps science going.
What are the features of a successful PhD student or postdoc?
I think the most important thing in science, but maybe even as a human being, is to keep your eyes, ears, and mind open to what you currently don’t understand, to what other people say, and how other people are – which helps you develop science, but also to understand the whole world better. For example, I started out as a chemist, but during my postdoc with Novartis I was dealing with High-Content Screening Data. I first found the data very difficult to analyse (and I still do!) – but this made me realize how important it is to look at drug discovery (and associated data) across its whole range, from chemistry, to biology and pharmacology, to physiology (etc.) All of those aspects are pieces of the big puzzle of drug discovery.
An open mindset helps in science, but also more generally to learn in life. I was for example quite influenced by my trips to India: I taught in Bangalore, at the Institute for Bioinformatics and Applied Biotechnology (IBAB) for about 10 years of my life, from my postdoc years until quite recently. This only happened since I replied to an email from Karthikeyan, from NCL Pune, a bit before, but it changed my life. If you are in India you notice for example what ‘go with the flow’ means – if 10,000 people want to exit Mumbai Central Station, then you just will be unable to enter. In life of course this means to not fight situations, but to deal with them as they come, to accept how things are.
The pictures below illustrate those years a bit – during my birthday celebration at the institute (the local tradition is to smear chocolate cake into the face of the lucky boy), as well as on the night train from Chennai to Bangalore. You get actually very good food from the vendors at the train stations, so next time you are in India, plan in a trip on the night train as well.


What is the most embarrassing thing you have done in the lab while doing experiments, e.g. explosions?
I started chemistry actually at home in my late teens – building rockets, and doing other science in what were ‘probably not entirely controlled conditions’. One time a chemical supplier sent me (to my home address!) the wrong delivery containing, among others, elementary sodium. It was too cumbersome, so apparently the supplier thought it a good idea to keep this delivery nonetheless, as a present. The label on the sodium, stored under liquid to protect from oxidation, said ‘do not bring in touch with water’, which… of course, as a teenage boy, is precisely what I did. Generation of sodium hydroxide, as well as hydrogen, and a decent explosion followed. I spent six weeks or so in hospital and nearly got blind as a result (bases cause much more serious damage to the cornea than acids), so I can confirm from empirical evidence: in case you find some elementary sodium at home, for whatever reason, do not bring into touch with water. I have realized that sometimes it is just fine to believe what one reads, and not everything needs to be proven in experiment.
What are your recommendations for a book, podcast, website, blog, YouTube channel or film?
As for resources, I can recommend for example Tom Mitchells’ ‘Machine Learning’ for an easy-to-read and very understandable introduction to the field; the blog ‘In The Pipeline’ for developments in and around the pharma area (https://www.science.org/blogs/pipeline), and TeachOpenCADD (https://volkamerlab.org/projects/teachopencadd/) from the Volkamer Lab for an hands-on introduction to the Computer-Aided Drug Discovery Field. I am also editing the Cambridge Cheminformatics Newsletter (http://www.drugdiscovery.net) with jobs and resources in the field which everyone is welcome to subscribe to, and where we organize regular events (also online) for those who are interested in the computational drug discovery field.
What do you think about the future of your field, AI in drug discovery, and how it applies to the advancement of medicinal chemistry as a whole?
We live in really exciting times currently, be it due to new therapeutic modalities, the amount of data being generated, or the computational and algorithmic power available. From the AI in rug Discovery angle, I would like to bring some realism to the debate – do we actually have the data available so that the field can live up to its expectations? I think this will be the case in some areas more than in others – for those interested in more details, the Cambridge Crystallographic Data Centre (CCDC) put one of my recent talks online which can be found here, on the topic “Artificial Intelligence in Drug Discovery – Where Are We Today, What Do We Need to Advance Further?”, which may be of interest to the reader here: https://www.youtube.com/watch?v=PrYIChWuUFI.