The Most Important Unasked (and Unanswered) Question in Drug Discovery: Is This Relevant?

Updated: Mar 10


Speaker Introduction - Dr Andreas Bender

Dr Andreas Bender is the leader of a research group in the Centre for Molecular Informatics at University of Cambridge, focusing on data-driven chemical and biological screenings for detecting mechanisms of action and adverse drug reactions, especially related to cancer.

Previously affiliated with the Lead Discovery Informatics group of Novartis in Cambridge/MA, Dr Bender now holds a leading position in the department of Clinical Pharmacology and Artificial Intelligence at AstraZeneca in Cambridge. He is also the co-founder of the Healx and PharmEnable startups which specialise in AI-driven small molecule discovery for various disease areas, including neurodegenerative, inflammation and rare diseases.

On his personal website, Dr. Bender is a smiling figure standing confidently in a warmly-lit library. A caption below the photograph introduces him as someone "committed to developing new life science data analysis methods". Alongside being a reader in Cambridge University’s Centre for Molecular Science Informatics and Director of Digital Life Sciences at Innovation Campus Berlin, Dr. Bender is also Associate Director of Computational Drug Safety at AstraZeneca. The fact that I recognised a drug company’s name is not the only thing that suggests we are most definitely in 2021: the conference is delivered via teams, and therefore my first impression of him is from his computer webcam.


Admittedly, as a literature student who has not academically engaged in science since the age of 16, I feel very much out of my depth. Perhaps this feeling of complete unfamiliarity may be at the root of general public apprehension towards taking the AstraZeneca vaccine.


Bender touches upon biased reporting in his talk, speaking about the issues of conducting a study when actively wishing for a certain outcome. However, Dr Bender is not dismissive of data that comes from this type of reporting— he explains to his audience how the results of these studies may be repurposed.


The Invivo Situation

Throughout the talk, Dr Bender emphasises the importance of the invivo situation when confronting disease and attaining cures. Where labels are a useful way of identifying something in theory, they are insufficient in reality due to their specific focus. We may understand that a certain compound binds to a certain protein, for instance. However, this data is insufficient due to conditional data lacking entirely from our studies. Bender explains that this lack of conditional data is one major issue that comes with treating certain diseases. The myriad of differences between two patients with the same disease - weight, phenotype, lifestyle, age and sex, for instance - uncover the need to take a look at the invivo situation— the bigger picture. Bender also touches upon co-morbidity: because of the overlap between both mental and physical illnesses, it is difficult to accurately label certain cases.


Labels in Drug Development

Labels, argues Bender, are probably not meaningful in the development of drugs. It can be commonly understood that a certain drug will cause a certain adverse reaction, and underlying causes may be entirely dismissed. How do the severity of these cases differ? Are there significant time differences in the time between taking the drug and having the reaction? Even the conditional differences in animal studies, where certain variables are easier to control, go unrecorded, despite having considerable effects on the outcome - animals tend to be less stressed when their handler is female, for example. What begins as a study on diseases and drugs morphs into a study of one compound amidst a rainbow of others. And while considerable data may be gained from such a study, Bender argues that its focus on unconditional variables makes it practically irrelevant.


This focus on significant yet small changes is attributed to pressure from the media. The looming fear in big Pharma is being overtaken by quicker, more nimble start-ups, and therefore it misses out on media attention and making fortunes from patents. Bender speaks of "the secret source": something that, as it appears, smaller companies are more prone to discover. The mysterious tone of the term captures interest while suggesting unattainability. “Hype brings money and fame,” Bender states. According to him, “reality is boring,” and indicative of the media’s complete disinterest when it comes to ordinary, yet practical change. Boasting major discoveries, however irrelevant to curing disease they may be, captures media attention, pressuring employees at Pharmaceutical Companies to show success no matter what. It is irrelevant how useful this data actually is in curing disease, statistical significance, ensuring clicks, and guiding the hand on what discoveries are more profitable. Admittedly, this major focus on profit was not surprising to me - the annual 11% increase in insulin prices come to mind in this prioritisation of profit over patient wellbeing.


However, information accessed from this type of labelling can be repurposed. Bender gives the example of both predicting the spread of HIV and treating patients with the disease. Understanding the relationships between certain chemicals and proteins allow us to determine what cocktail of medication a patient may get, as well as the mode of action from there. Although the method used to retrieve this data was not useful, that does not mean that the data gathered cannot be used in a positive way.


Bender also recognises the abilities of AI in image and speech recognition. In image recognition, AI works on a pixel-based basis, detecting edges on a picture, and adding a largely unconditional label. As mentioned earlier in reference to investigating proteins and compounds, these are fewer used when we do not understand the conditional labels in drug discovery. Quite often the masses of data accumulated with AI uses proxy systems - fast and cheap, but unrepresentative of the bigger picture. Masses of data are generated simply because it is cheap to do so, and we are saddled with a host of figures, largely irrelevant when it comes to real-world uses.


Further issues arise with the creation of what Dr Bender describes as a ‘feedback loop’ - newer technology getting used simply because it is new, and not because it is relevant to this situation. Research becomes so far-removed from reality, living to serve the short attention span of the media, that the question”is this relevant” is asked less and less. The data we are presented with is significant, detailed - but little more than that. Its role in the world is simply to capture your attention when browsing the internet, as opposed to curing diseases.


What Bender argues in relation to data collection reveals not only an issue with the Pharmaceutical Industry, but also in the world as a whole - that the sole purpose of all of this work is to capture our short attention span rather than cure our diseases. I am reminded of headlines taken out of context, photos and soundbites spliced in an incriminating manner, a shameless bid to gather clicks as opposed to meaningful content.

We already have a proposed solution to many of the issues brought up by Bender - focusing on the larger situation, attempting to uncover the underlying causes - and what matters now is investigating this instead of the best way to capture the world’s attention.

To learn more about Dr. Andreas Bender and his work,

visit Dr Bender’s personal website.

Related Posts

See All
 
OUR PARTNERS

UCL

STATISTICS SOCIETY

  • Facebook
  • Instagram
  • newsletter-icon
145718758_111103884288370_64813884565358

UCL

SELCS

  • Facebook
  • Instagram
The%20Future%20of%20Communication_edited

SUSTAINABLE UCL

  • Facebook
  • Instagram
71184778_2227947534000972_27060017277477

Subscribe to our Newsletter!

  • Instagram
  • LinkedIn
  • Facebook

©2021 by enlightment.