The fundamental issue in cancer research has always been one of discrimination. Cancerous cells and healthy cells are molecularly similar; what makes them different is that in cancer cells there is an aberration in how certain switches have been turned on, leading to abnormal growth. Identifying these differences used to require years of searching for minute patterns within sample cells collected from cancer patients.
Advances in AI mean that what can now be achieved is a far cry from this process. Genomic data compiled from tens of thousands of cancer sequences can be fed into AI systems to detect the genomic patterns in cancerous tissue alone. These patterns will be far more sophisticated than biomarker systems in earlier iterations of precision medicine and would allow the AI to differentiate cancerous and healthy tissue at the level of gene regulation.
Using such patterns, many novel ways to approach cancer become available, such as designing personalized cancer vaccines using AI.
Both Moderna and Merck are already testing such technology at the late stage of their clinical trials, based on the same messenger RNA platform used for the production of the COVID-19 vaccine. Furthermore, artificial intelligence is assisting in designing more intelligent CAR T-cells, which can continue working within the immunosuppressive environment of cancer by relying on cancer-specific signals. Finally, at the very beginning of the drug development process, artificial intelligence enables analyzing genomics and imagery data to identify cancers even decades earlier when symptoms begin to show up.
What do we currently use to treat cancer?
The current state of the art consists of scientists discovering natural targets existing on or inside the tumor cell (proteins, enzymes, and receptors) and designing drugs against those targets. It is expensive, time-consuming, and severely limiting. It is because apart from existing in tumor cells, natural targets also occur in healthy cells. Therefore, any activation of the immune system by drugs occurs simultaneously in other cells, resulting in a storm-like effect detrimental to patient's health.
It is why we currently try to lower the dosage in an attempt to mitigate that problem. Unfortunately, lowering the dose means lowering efficacy, increasing the risk of relapse. Once cancer recurs, it is likely to be resistant to our drugs since it was granted enough time to undergo mutations and adapt to them.
In case of lung cancer, which is responsible for 1.8 million annual deaths worldwide and is therefore considered one of the deadliest cancers, we managed to increase its 5-year survival rate twofold during the last twenty years. Still, it means that 70% of people with such cancer will pass away within five years after their diagnosis.
AI-driven cancer bioengineering explained
The use cases for artificial intelligence have proven to be far more consequential than just building impressive chatbots. Sure, applications like helping radiologists identify abnormalities in images quickly or mining databases for potential drug repurposing opportunities were a good start. However, something else is needed, something radically different.
Here, the metaphorical description that scientists began using recently would compare this use case for artificial intelligence to what AlphaFold has achieved for proteins. AlphaFold has not discovered proteins themselves but rather the principles underlying how they fold, providing a systematic understanding of protein structures.
AI-driven bioengineering of cancer does the same thing for the rules guiding the genetics of cancer. The result is an ability to write code that runs within a cancerous cell as precisely as no biological markers ever allowed. This is not about reading the code but rather rewriting it in ways previously unimaginable.
For the code to be delivered into the cancerous cell, there must be yet another leap forward made – the one in the method. Namely, for the artificially created genetic load to get to the cancer cell, it must survive immune attacks along the way. Scientists believe they found the solution to this problem in the form of lipid nanoparticles used during the development of vaccines against the COVID-19 pandemic.
The programs for the pandemic demonstrated that these nanoparticles can be used successfully to deliver mRNA payloads into human cells in a safe manner. Now, bioengineers are taking advantage of this discovery, preparing this mechanism to transport cancer-specific genetic loads. Of course, thanks to AI algorithms decoding vast screens and libraries of compounds.
China is already ahead
Yet none of this will matter if America does not take it seriously. China has placed biotechnology at the forefront of its strategic agenda, investing state money in biotech startups, reducing regulatory review timelines, and posing a credible challenge to America's supremacy in the field. In just the first half of last year, the pharmaceutical sector invested $48.5 billion in Chinese biotech ventures, surpassing the total investments of all of 2024. The United States continues to direct its venture capital heavily towards AI in the software-only sense. Last year, artificial intelligence companies raised over $200 billion in venture capital investments; this amount is 50% of all venture capital funding. Biopharma raised around $26 billion.
This discrepancy is not merely a market inefficiency. It is an example of a blind spot that prevents one from understanding that the most significant use of artificial intelligence in the next decade might not involve making software smarter but rather altering the tangible reality we inhabit, thereby extending biology from a scientific discipline into an engineering discipline through cellular programmability.
What needs to happen for the US to lead on cancer treatment
In order for the US to lead on the next generation of cancer treatment, there should be a national fund specifically dedicated to investing in biotechnology startups. This is not just the ARPA-H, which remains a largely university-focused effort, but an initiative that takes a more direct approach in making capital available to the platform companies creating the next-generation technologies, keeping the intellectual property in the US.
It is also important for institutional investors and VCs to get involved in biotech beyond their focus on software AI and start considering whether a company working on a technology that will allow cells to fight off cancer is worth the same attention as a new large language model.
Finally, the FDA's new accelerated approval framework needs to be expanded to include biological platforms that do not depend on individual assets, so that those companies developing next-gen biotech products are not left waiting a decade before gaining regulatory clarity.
The science is ready. The US needs serious and sustained investment in America's future precision medicine technologies to make sure cancer programs itself.
Post a Comment