Meet Lisandra. She’s a Machine Intelligence Specialist.

Not as cool as vultures and bats, but still cool enough

Tech and digital are my fields of work; everything I do involves AI. As a Specialist in the Machine Intelligence Department in Novo Nordisk, I work with various AI algorithms, particularly graph deep learning, graph data science, and generative AI. I use these to discover new drug targets whilst striving to understand their potential modes of action. I also work with patient-level data to identify disease subtypes, aiding in patient stratification because doing so can enable more targeted drug treatments and enhance disease understanding.

More generally, I work with what we call 'biomedical knowledge representation' approaches. This essentially involves working with complex high-dimensional data and highly relational data, transforming it into lower-dimensional – or simpler – data with the aim of facilitating the answering of the variety of questions we might have.

Detente, sombra de mi bien esquivo…

✎ Back in my younger days, I wrote and self-published four novels and two poetry collections. I still write poetry now and then, whenever inspiration strikes.

Joining

It began with microbes that can kill you in a matter of days

I was a Marie Curie Fellow in the Institute of Microbiology and Infection at the University of Birmingham, UK, when I decided to leave academia. This was around the time when Covid hit, and it changed the research landscape for a lot of people in the UK – adding the issue of Brexit. As a Fellow, I was modelling the virulence networks of antimicrobial-resistant pathogens. I’ve always thought drug resistance to be a super interesting topic, so when I found that some patients also could become resistant to their current treatments, I started thinking of humans as an interesting model to study too. Not as cool as vultures and bats, or microbes that can kill you in a matter of days, but still cool enough.

So, when I saw an ad for a Senior Scientist position at the Novo Nordisk Research Centre Oxford, I thought it was worth a shot. The position was in the Computational Biology Department, and I was going to be working with graphs and other machine-learning approaches to discover drug targets. It sounded like the natural next step in my career to keep developing my computational biology, bioinformatics and machine learning skills.

Luckily, they weren’t Swedish miles

🏃‍♀ I accidentally signed up to a half marathon, because I confused miles with kilometres.

Novo Nordisk always seemed like a nice company to work for. The Danes have a good work culture, so I assumed that working in a Danish company would have a similarly good culture – something that is hard to find in other large American or European companies. After joining, it surprised me how much of the Danish culture the company retains in the UK offices – even after growing to be such a global workplace. It still has a flat structure, and you can easily interact with senior colleagues, even with members of the leadership team. That is pretty unique.

Today I am located at the Novo Nordisk Research Centre Oxford, and I've been pleasantly surprised by all the opportunities for research that we have because of being in Oxford. A special mention goes to the brilliant Oxford and Cambridge University students who do internships with us.

Work

A big friendly GiaNNt

I’m very lucky, I feel like all my projects are cool. What I really enjoy is finding hidden patterns in the data, patterns that might help us understand how this particular gene or set of genes is related to a particular aspect of a disease. But if I were to mention one project in particular, it would be the development of an in-house biomedical Knowledge Graph, bringing together public and in-house data. Modelling relational data as a graph is so much fun. We called this knowledge graph “GiaNNt” which stands for “Graph-based Identification and Annotation of NN targets” and we used it exactly for what its name says. Based on it, I developed projects focusing on identifying targets for atherosclerosis, obesity, NASH, and insulin resistance. Specifically, I found insulin resistance to be the most interesting. I developed it together with my colleagues by testing different graph neural networks, other simpler graph topology, and machine learning approaches. It was a lot of learning, not only technical, but also biological.

Across projects, I mainly work with other machine learning scientists like myself – in my team, across departments or with externals – but I also work with biologists from Global Drug Discovery, such as scientists working in the lab, and subject matter experts from the company’s other therapy areas. And of course, university students doing research internships or master projects.

Technologies

You say data, I say graph

I mainly work with Knowledge Graphs. Taking highly relational data, such as the biomedical data from various different databases, and putting it into a graph where all the data is harmonized in a single place, and can be used for querying or as input for graph algorithms. I also work with unstructured data using Large Language Models, Natural Language Processing, and other text mining techniques.

I recently started working with patient level data to model patient disease progression, predict disease risk, and understand disease endotypes, i.e., understanding the different physiological mechanisms that lead to a subtype of a disease. I do all these using high-performance computational environments, using mainly Python and Shell. Although, my R is getting a bit rusty.

Communities

Machines take me by surprise with great frequency

I benefit a lot from being involved in several communities of practice. Of course, those related to Knowledge Graphs, Data Science, and Large Language Models. I am part of the Alan Turing Knowledge Graph Interest Group that facilitates research and innovation in this field. And I’m also part of a Novo Nordisk Knowledge Graph Community of Practice together with other people from NN Development and NN Digital & Research Intelligence. Here, we share and discuss state of the art methodologies and relevant articles.

Learning milestones

␥ Publishing my first git repository as part of my PhD, and making the code available for anyone to use. 
␥ Building the first inhouse Knowledge Graph for R&ED in Novo Nordisk. 
␥ When I transitioned from the Computational Biology Dept to the Machine Intelligence Dept and started working more with Large Language Models. 
␥ Getting my first PhD student, who is always challenging me to learn more about certain state of the art machine learning model architectures.