In the wake of new programs that produce artwork derived from existing media, ChatGPT, and even algorithms that can predict protein folding, it is evident that the age of artificial intelligence (AI) is upon us. In many cases, the AI programs and tools are far more advanced than we have previously seen, to the point where humanity can derive great benefit from AI while fearing how it may affect our society and livelihoods. While it is unlikely that we will be subjugated by our new robot overlords, it is still important to explore what has been done and remains possible through AI, and our considerations for its ethical usage.
Defining Artificial Intelligence
Aside from the science fiction stories of killer sentient robots (that have since become more or less reality, minus the killing), artificial intelligence is a branch of computer science that tries to simulate higher intelligence through machines and software. With the growing use of computers in modeling, proteomics, genomics, and more, the next logical evolution was to use the computing power to drive more efficient predictive algorithms to focus our resources on more fruitful avenues of research. I recall during the days of my undergraduate and early graduate education that idle computers and even gaming consoles were being used to provide computing power for huge projects like SETI@Home and Folding@home, which was a precursor to the modern protein folding algorithms.
Although there are higher philosophical goals to generating an advanced artificial intelligence, most of the modern applications are focused on problem-solving and learning to refine predictive models. A lot of the principles in my computer science classes involved information recall, sorting, and storage, and as computers and programming languages evolved, so too did the strategies to leverage that increased computing power. For example, you may have heard terms like “neural network” and “machine learning,” which refers to the modeling strategies employed to organize and decipher huge datasets similarly to how neurons in complex nervous systems communicate with each other. The goal of these computational strategies is to improve predictive algorithms that are trained on prior experimental data in order to drive new research directions in a manner that is far more efficient than prior “shotgun” type approaches. In focusing research directions, the AI and other computational helpers can reduce the resources used in terms of time, personnel, and reagents.
Applications of Artificial Intelligence
In addition to better algorithms to target content to users on social media and sites such as YouTube, AI is heavily leveraged in modern times to drive scientific discovery. In particular, AI has changed the landscape of biomedical research, mostly for the better. There are various logistical applications for AI, which includes supply chain optimization and drug pricing, but for the bench and computational scientists, the motivation for using AI is in the potent predictive algorithms that drive drug and biomarker discovery. 1 For example, machine learning was employed to discover a new antibiotic against a notoriously resistant bacterial pathogen. By training a neural network with experimental data comprising molecules that were found to inhibit the growth of the superbug Acinetobacter baumannii, scientists were able to discover abaucin, a structurally new molecule with effective performance against the pathogen that was derived from the AI-driven analysis. 2 There are many other applications of AI, including some that I didn’t even know were subjects of interest, such as using the brain’s ability to remember faces to improve facial recognition systems, enhancing our understanding of gene activation and infectious diseases, and even a program that can accurately identify species in the wild. The only limit thus far seems to be the imagination of the programmer, but once the algorithm is released to do its stated function, the discoveries can be quite impactful.
Artificial Intelligence Fighting for Market Share
Every year, it seems like dozens of new biotech startups emerge with an exciting new approach to improving the human condition, and others explode into the limelight as they develop their big breakthrough. Some of them persist, and some are absorbed into existing big pharma companies, but the fact is that these companies can each acquire hundreds of millions of dollars in funding, leading to billions in revenue and profit. Among the perceived best companies of 2024 is Cellarity, which uses proprietary AI models to develop novel drug candidates across a variety of diseases.
The potential of AI to improve global human health (and all the money they can make from doing that) is certainly not to be ignored by big business. In January 2024, we have already seen a partnership between Isomorphic Labs, an AI-based drug discovery entity, and both Eli Lilly and Novartis, through transactions that total about $3 billion if performance milestones are met. Isomorphic will receive upfront cash from the big pharma companies but can generate billions should they generate viable new medications to treat a broad range of diseases. Similarly, Abbvie and BigHat Biosciences announced a research collaboration in December 2023, hoping to leverage BigHat’s AI-based Millner platform to drive therapeutic antibody discovery and engineering. This deal gives BigHat some $30 million up front and the capacity to earn $325 million in performance milestones plus future partnerships.
It may not be only the usual suspects in biopharma getting into the act, as ByteDance, the parent company of the popular social media platform TikTok, is starting to recruit talent in computational biology and other disciplines to set up AI teams for drug design and science. Although it isn’t totally clear why a social media company is getting involved in AI-driven bioscience, experts suggest that TikTok data could be used for large-scale hypothesis generation based on the content produced by its many users, especially those who are interested in health, science, and well-being focused content.
Ethical Considerations for Artificial Intelligence
Because AI models are reliant on their training datasets, the availability of accurate and vetted information via those datasets is extremely important. Also, the accuracy of predictive models is sometimes in question, and even the best models must be validated through experimentation and clinical testing. 1
According to at least one forecast, business spending on AI is expected to hit $110 billion globally in 2024, with many companies already adopting multiple AI systems to leverage this growing technology. And because AI is trained on datasets, there will be discussions on how and which data can be accessed, pursuant to privacy and confidentiality laws. There are societal ramifications in the increased usage of AI, similarly to the trepidation when industries began to automate many tasks, but the positive outlook is that by automating certain tasks or analyses, we can focus human talents towards other aspects of our society.
Not unlike the recent strife in Hollywood over AI usage and its impact on the workforce, particularly for writers (who could find themselves forced out with ChatGPT generated scripts) and actors (who could find their likeness and image rights usurped by AI art and animation generators), AI permeating nearly all aspects of our society could generate biases or discrimination in many forms. To that end, various governments and organizations including UNESCO are working to better understand AI and to properly regulate it with ethics and maximum benefit to humanity in mind.
Overall, AI, as with many other scientific innovations, has strong potential to change humanity for the better if used wisely and responsibly. With mindful regulation and personal vigilance, AI can continue to benefit science and society without threatening our way of life through invasion of privacy and various biases generated by its modeling.
Artificial Intelligence Enhances Antibody Engineering
Dr. Hao Cheng, PhD, of Yurogen joins BioChat to discuss the many considerations that go into engineering a more effective antibody reagent for research and therapeutic use. Dr. Cheng and Yurogen leverage the power of artificial intelligence to improve binding and stability of the antibody among many other applications of this amazing tool. Listen below and click here to browse our archives!
References
- Bhardwaj A, Kishore S & Pandey DK. (2022) “Artificial Intelligence in Biological Sciences.” Life (Basel) 12(9):1430 (Epub).
- Liu et al. (2023) “Deep learning-guided discovery of an antibiotic targeting Acinetobacter baumannii.” Nature Chemical Biology 19:1342-1350.
