New potential cancer drugs and where to find them
Cancer research generates massive amounts of data, but traditional tools often fail to fully harness their potential. How can we unlock this data to provide better treatments for cancer patients? PhD candidate Marina Gorostiola González explored this by using advanced data analysis techniques to guide drug discovery.
Gorostiola González’s interest in using data analysis techniques in pharmacological research began almost by accident. She was looking for a final dissertation project on data analysis and drug modeling for clinical trials but found no ongoing trials at the time. Instead, she worked with virtual patients—a method she grew to appreciate for its ability to save time, resources, and prevent harm to actual patients or animals.
Analysing large amounts of data in ways we couldn’t before
During her PhD research here at the Leiden Academic Centre of Drug Research, she focused on how to best use computational tools to develop new and better drugs for cancer patients. ‘We now have access to incredible amounts of data,’ Gorostiola González explains. ‘Computational tools allow us to analyse it in ways we couldn’t before, uncovering insights we might never even have considered.’
‘We now have access to incredible amounts of data’
Computational tools have long been used in drug discovery, but Gorostiola González aimed to apply them specifically to cancer research, an area that had not yet fully explored this potential due to limited data. Although she didn’t find a cancer cure, Gorostiola González believes her approach will contribute to developing better therapies in the future.
‘Without experimental data, computational analysis is impossible’
Her research focused on membrane proteins, which sit on cell surfaces. These proteins serve as entry points for signals and substances and hold great potential for cancer therapies. Unlike kinases, the current go-to targets in cancer treatment, membrane proteins had not been extensively studied in this context. Gorostiola González explains that they’ve been overlooked because they are challenging to work with, both experimentally and computationally. However, new advances in experimental methods have now made gathering data on these proteins more feasible, enabling deeper computational analysis. ‘Without experimental data, computational analysis is impossible.’
The three-step approach to new cancer drugs
To find new cancer drug targets, Gorostiola González followed three steps, each requiring different types of data. First, she predicted new potential targets. For this, she used so-called omics data: large-scale biological data that provides insights into the complete set of genes, proteins, metabolites or other molecular levels. ‘For instance, we might look at the DNA of 1,000 patients with a specific cancer type and compare it to the DNA of 1,000 healthy individuals. A model can be trained to identify changes in the DNA that predict whether someone has this type of cancer. These changes might be interesting as new drug targets.’
By integrating various types of data, we can build a comprehensive picture of what’s happening in the body.
The second step is refining the targets. Structural data, which shows the physical structures of proteins, DNA or other cellular components, helped Gorostiola González refine the potential targets. ‘We examined specific alterations in the potential targets that make them better candidates for therapy. If a mutation is unique to cancer, we target it without effecting normal tissue, reducing side effects.’ Structural data is often publicly available, and models from the AI tool AlphaFold help fill in gaps when experimental structures are unavailable.
Finally, Gorostiola González used biochemical data to predict which drugs would bind effectively to the refined targets. ‘This gives some final hints towards promising new drugs.’ By integrating various types of data, we can build a comprehensive picture of what’s happening in the body, without the need for extensive experimental and clinical experiments.
The value of collaboration and sharing existing data
Gorostiola González concludes that computational tools have already proven useful in this context but that challenges remain. For example, greater collaboration across disciplines could lead to significant advances. She highlights that combining experimental data with computational models allows scientists to refine drug predictions and improve the chances of success.
However, a significant obstacle persists: the lack of data in oncology. While models and computational power have advanced rapidly, the absence of comprehensive data continues to limit breakthroughs. Gorostiola González points out that sharing existing data, including results from failed experiments, could be invaluable.
Gorostiola González concludes by underscoring the importance of computational tools in preselecting promising experimental options. ‘Instead of testing 1,000 compounds, you can narrow it down to 50. While there’s no guarantee that all 50 will work, it still saves time and resources.’
PhD defense
Marina Gorostiola González defended her dissertation, Getting personal: advancing personalized oncology through computational analysis of membrane proteins, on 24 January 2025 at the Academy Building. She received her doctorate cum laude. Her supervisors were professors Gerard van Westen, Laura Heitman and Adriaan IJzerman.