The Challenge of Chemical Space and the Role of Artificial Intelligence
Why Small Molecules Matter
Small organic molecules are the workhorses of modern chemistry. With molecular weights typically ranging from 50 to 500 Daltons, they are built from the familiar atoms of carbon, hydrogen, nitrogen, oxygen, and occasionally sulfur, halogens, or phosphorus. Despite this simple alphabet, they exhibit an extraordinary range of chemical and biological functions.
They underpin therapeutics—from aspirin to antibiotics to advanced oncology drugs. They play a pivotal role in agriculture, as herbicides, insecticides, and growth regulators. They enable materials science, appearing in dyes, polymers, and semiconductors. They even influence defense and security, in detection systems and protective coatings.
The breadth of applications is astonishing. Yet the same diversity that makes small molecules so powerful also makes them difficult to systematically discover and optimize.
The Vastness of Chemical Space
At the heart of this challenge lies the immensity of chemical space. Chemists estimate that the number of possible small molecules lies somewhere between 10⁶⁰ and 10⁸⁰. This is an unfathomably large number, dwarfing even the number of stars in the observable universe.
Why does this matter? Because within this vast space are molecules that could solve pressing problems in medicine, sustainability, and technology. Somewhere out there may exist a compound that cures a resistant infection, captures CO₂ from the atmosphere more efficiently, or enables a new type of electronic device.
But the scale is both opportunity and obstacle. With current laboratory capabilities, there is no way to enumerate, synthesize, and test every possibility. Even brute-force screening of millions of compounds represents only a pinprick in the full landscape of what could be discovered.
The Traditional Drug Discovery Cycle
When focusing on drug discovery, the process often follows a prototypical design cycle:
Design a set of molecules that might exhibit desirable features such as potency or solubility.
Synthesize those molecules through organic chemistry in the lab.
Test them experimentally in assays, often starting in vitro before moving into biological systems.
Refine the design based on results, and repeat the cycle.
This cycle is iterative and costly. A single loop can take weeks or months, and dozens of cycles may be required. Hundreds or thousands of compounds may be synthesized before one candidate emerges that merits pre-clinical and eventually clinical development.
The result is that time, cost, and uncertainty are immense. Entire projects can collapse because the right compound cannot be found quickly enough.
How AI Enters the Scene
For decades, researchers have turned to computation to ease this burden. In the 1960s, the field of quantitative structure–activity relationships (QSAR) emerged, attempting to correlate molecular structure with biological function. Later came virtual screening, where computational models are trained to predict the activity of hypothetical molecules before they are ever made in the lab.
These methods save enormous resources by filtering out poor candidates early. Instead of synthesizing thousands of molecules, researchers can focus on a smaller, more promising subset.
But even with these tools, one bottleneck remains: the need for human chemists to propose candidate molecules in the first place. Models can rank options but rarely create them. The search is still constrained by imagination and prior knowledge.
Conclusion: The Need for a New Paradigm
The immensity of chemical space, combined with the slow grind of the design cycle, demands a new approach. Traditional AI tools help prioritize, but they do not invent. To unlock the next generation of therapeutics, materials, and agricultural solutions, we need systems that can go beyond screening.
This is where generative AI comes into play: flipping the problem on its head by asking, if I want this function, what molecule should I build?
In the next article, we will explore how generative AI is transforming molecular discovery—shaping not only what molecules are proposed but also how we might make them in the laboratory.
