Better. Cheaper. Faster. The much-sought trinity of drug discovery. Attend any conference in the pharma sector, and you will repeatedly hear how artificial general intelligence or AGI in drug discovery is the solution to fast-track medicinal chemistry and produce better, cheaper drugs in next to no time.
Here’s the thing, though: since the machine learning explosion, drug discovery has slowed down. It has also become more expensive, and there isn’t much evidence that it is all that much better yet either. So, what is happening? Is the AGI in drug discovery promise all hype? Are we just not quite there yet? Or is there something else going on?
Is AGI in drug Discovery all hype?
For the hype question, the answer is categorically no. The lack of apparent progress is, for the most part, a matter of relative timelines. Getting new drugs to market takes time, which means that even though the ML (machine learning) explosion has been building for a while, developments that have any significant effect on actual time-to-market have not yet come to fruition. In the meantime, more stringent regulatory requirements, increasingly complex clinical trial designs, the need for more comprehensive safety and efficacy data, and all the attendant costs have extended the time it takes to launch any new drug1. To date, these factors have muted any ML-driven gains, but that doesn’t mean this will always be the case.
AI-enabled drug discovery is already demonstrating significantly reduced exploratory research phase timelines. The lack of apparent progress makes for the occasional good headline, but the trickle of data coming from AI-facilitated programs (scant due to the industry-wide tendency to keep cards close to the chest) gives us good reason to be hopeful.
Many AI-facilitated programs are completing their discovery and preclinical phases in four years, much shorter than the historical industry standard of five to six years1. In some cases, this timeframe is shorter still, such as Exscientia’s AI-generated small-molecule drug candidate DSP-1181. Intended for treating obsessive-compulsive disorder, DSP-1181 moved from exploratory research to a Phase 1 clinical trial in less than 12 months2. These results are very positive, but they are only the beginning. We can do much more now to ensure AI-enabled algorithmic solutions become game-changers in maximizing the return on research dollar investment.
What are we hoping to achieve using AGI?
The kind of impact we are all hoping for and working towards can only be achieved by addressing the major challenges currently faced in the use of AGI in drug discovery and the journey to getting an effective drug to market. The success of AI models heavily depends on the quality, quantity, and connectedness of available data, which is (surprisingly, given years of “big data” focus) still a significant challenge in biopharma. Even when we do get good enough data, the “black box” nature of many AI models can lead to challenges in interpretation and potential biases in the results. Also, the bigger and more impenetrable the models get, the more difficult it becomes for the still-vital humans in the loop to understand and communicate the rationale behind the suggested candidates. This, in turn, makes integrating AI tools into established drug discovery workflows and gaining trust from stakeholders even more challenging.
These challenges mean a lot of the work to date involving AGI in drug discovery tends towards tinkering - staying safely in the hottest spaces, improving efficiency but still sticking as closely as possible to the traditional process. With its expected capability to understand, learn, and apply knowledge like an omniscient genius, if artificial general or super intelligence (i.e. AI at or beyond human-level intelligence) is created, most of us accept that this would most likely mean a plethora of novel hypotheses, experimental designs, processes and associated drug candidate suggestions developing with minimal human intervention. However, while we kick our heels in the AGI waiting room, we needn’t be idle. After all, if we are open to the idea that AGI can truly crack the “Better, Cheaper, Faster” problem, why not try now?
Is now the time to incorporate AGI?
Solving the data problem by filling in gaps and finding connections requires providing real context. The traditional drug discovery process often involves sifting through vast amounts of biological, chemical, and clinical data, which are not always of the best quality or completeness. This process is time-consuming and prone to missing crucial links between disparate pieces of information.
Algorithmic solutions can extract, compile, parse, and contextualize these data sets. By integrating these various data sets, algorithms within existing model types can create a complex relational context, providing researchers with a holistic view of potential drug candidates and their data networks. Such solutions go beyond simple data analysis. They enable the identification and harnessing of previously overlooked relationships between compounds, biological targets, and diseases, between the entirety of what is known, what has been explored, and what could be.
How does visualization fit in with AI modeling?
Equally, you deal with the black box problem by shining a light. Most humans tend to understand something when they can see it. This means visualizations, which can show the multi-dimensional data space and how new data fits within and shifts the relationships within the space. These visualizations, therefore, become the vital conduit between AI modeling and human understanding and decision-making.
By translating complex, abstract data into accessible visualizations, the gap is bridged between machine logic and human intuition. This approach not only enhances transparency but also aids in identifying patterns, outliers, and trends. Moreover, interactive visualizations, which allow users to manipulate data parameters and see immediate changes, empower decision-makers to explore “what-if” scenarios and better understand the implications of different choices.
This visual approach is especially crucial in biopharma, where understanding the underlying model logic can lead to more informed, ethical, and accountable decision-making. Due to the clarity they provide, visualizations foster greater trust and collaboration between human and machine, ensuring that AI tools are not just black boxes but transparent and interpretable aids to human expertise.
What does this mean for researchers today?
AGI in drug discovery used as a facilitator for development is still in its infancy. While we may sometimes need to take a breath and remember the time frames we’re dealing with, that is no reason not to be excited or to accept relatively minor process improvements. Through improving, tailoring, and contextualizing the available data, while using visualization to bridge the understanding gaps between humans and machine, we can help usher in a paradigm shift in innovation and efficiency in drug discovery. Irrespective of how long we may have to wait for AGI to get here, we can begin to be better, cheaper, and faster sooner.
Accencio, LLC is a technology company driving innovation at the nexus of information and design. Our cloud-based subscription products provide fully visualized, data-rich interactive landscapes that accelerate and enhance research, discovery, and commercialization across multiple science-based industries and beyond. Contact us today to understand how your research can be contextualized and enhanced using our flagship product IP-GeoScape.
1Jayatunga, M. K., Xie, W., Ruder, L., Schulze, U., & Meier, C. (2022). AI in small-molecule drug discovery: a coming wave? Nature Reviews Drug Discovery, 21(3), 175–176. https://doi.org/10.1038/d41573-022-00025-1
2Kirkpatrick, P. (2022). Artificial intelligence makes a splash in small-molecule drug discovery. Biopharma Dealmakers. https://doi.org/10.1038/d43747-022-00104-7
By
·
4 minute read
Better. Cheaper. Faster. The much-sought trinity of drug discovery. Attend any conference in the pharma sector, and you will repeatedly hear how artificial general intelligence or AGI in drug discovery is the solution to fast-track medicinal chemistry and produce better, cheaper drugs in next to no time.