From SMILES to Medicine: Understanding Where My Role Ends — and Why That’s Enough

Here, CO and OC indicate methoxy groups attached to aromatic rings, c1ccc2…c1 encodes the fused aromatic/heteroaromatic ring system, [nH] represents a protonated nitrogen in the benzimidazole ring, and S(=O) captures the sulfoxide linkage that is central to omeprazole’s chemistry and mechanism of action. Like all SMILES strings, this is one of several equivalent ways to write the same molecule; different canonicalization rules or software may produce a slightly different—but chemically identical—string.

When people hear that I’m building a cancer drug discovery engine, a common assumption follows:

“So… you’re making cancer drugs?”

The honest answer is both yes and no. Let me explain it. The question makes sense. For decades, drug discovery has been presented as a single, monolithic process, a black box where ideas go in, and years later, if you’re lucky, a medicine comes out. In reality, it has never worked that way.

Drug discovery is a chain of specialized roles, tightly coupled but fundamentally distinct. Computational scientists explore molecular possibilities. Chemists turn viable structures into physical matter. Biologists test effects in living systems. Clinicians translate those results into therapies. Regulators guard safety. None of these roles can stand alone. And when they are poorly connected, progress slows to a crawl.

The traditional model reflects this disconnect. It often takes 10 to 15 years from initial hypothesis to approved drug, not because people are careless or inefficient, but because too much effort is spent testing the wrong ideas too late. Chemistry begins before exploration is thorough. Biology begins before molecular understanding is solid. Wet labs absorb enormous costs validating candidates that should never have reached them.

The result is a discovery pipeline that is both expensive and exclusive, accessible only to large organizations with deep pockets, long timelines, and tolerance for massive attrition.

What I’m trying to do is not to replace any of these roles, but to tighten the coupling between them.

Computation, chemistry, and biology should not operate as isolated silos. They should form a feedback loop — one that eliminates weak candidates earlier, preserves context instead of discarding it, and allows human expertise to focus where it matters most.

If exploration can be done more deeply before synthesis, and if uncertainty can be handled intelligently rather than ignored, then the downstream work becomes more efficient, not faster in a reckless way, but shorter in an honest one.

That’s where my role sits.

I’m not shortening the journey by skipping steps. I’m trying to shorten it by making each step more informed.

And if that works, even partially, it opens the door to something important: democratization.

Not democratization in the sense of making drugs easy, but in the sense of making early discovery accessible to small labs, startups, and individual researchers who currently can’t afford a decade of blind digging.

That’s the context in which everything else in this post should be read.

Over time, I’ve learned that drug discovery is not a single act. It’s a long relay race. And understanding which leg of the race you’re running is just as important as running it well.

This post is my attempt to clarify that for myself as much as for others.

The Part I Understand Deeply

The early phase of drug discovery happens before a single molecule is synthesized. This is where modern computation plays its most critical role.

This is the domain I work in.

It includes:

• Docking — exploring how millions of candidate molecules might bind to a protein target

• Molecular Dynamics (MD) — testing whether those bindings are stable over time

• Binding stability & selectivity — asking not just does it bind, but does it bind better than it binds elsewhere

• IC₅₀ modeling — estimating potency and dose–response behavior

• ADMET prediction — early signals of absorption, toxicity, metabolism, and clearance

• Synthetic accessibility scoring — a reality check: is this even makeable?

This phase eliminates the vast majority of bad ideas. And it must.

Because the chemical universe is unimaginably large. No lab on Earth can afford to physically test even a tiny fraction of it.

What I’ve built with QuantumCURE lives here: the exploration and reduction of chemical possibility space. This alone can remove 90–99% of candidates before any wet chemistry begins. That’s not trivial. That’s survival. That is massive and vital.

What a SMILES String Really Is (and Isn’t)

A SMILES string is often misunderstood.

It is not a drug. It is not a recipe. It is not a synthesis plan. A SMILES string is a graph-level description of a molecule, atoms, bonds, and connectivity. It tells us what a molecule would look like if it existed. For example:

SMILES stands for Simplified Molecular Input Line Entry System, and it is a compact text notation used to describe the structure of a molecule in a way computers can easily read and manipulate. For example, the SMILES string CC(=O)O represents acetic acid: the first C is a carbon atom (a methyl group), the second C is another carbon connected to it, (=O) indicates that this second carbon is double-bonded to an oxygen atom, and the final O represents a hydroxyl oxygen bonded to that same carbon. In SMILES, atoms are written as their chemical symbols, single bonds are implied by adjacency, double and triple bonds are shown with = and #, parentheses indicate branching, and special symbols can encode ring closures and stereochemistry. In short, a SMILES string is a linear, text-based way of encoding a molecule’s atomic connectivity—not how to make it, but what it would look like if it existed.

It does not tell us:

• how to make it

• whether it’s stable

• whether it’s toxic

• whether it survives metabolism

• whether it can even exist in a flask

In other words:

A SMILES string is an invitation, not a commitment.

Most invitations should be declined.

My work is about declining the right ones early.

Here is the Part I Do Not Control — and Respect Deeply

After computational discovery comes a world entirely different.

This is where chemistry becomes physical.

At this stage, expert chemists ask questions like:

• Can this molecule actually be synthesized?

• How many steps will it take?

• Are the intermediates stable?

• Are the reagents known and available?

• Is this scalable or just a lab curiosity?

This process, called retrosynthetic analysis, is both an art and a science. There are AI tools that help. There are machines that execute reactions. But there is no magic button that turns a SMILES string into a pill.

What exists instead is:

• automated synthesizers that execute known chemistry

• flow chemistry systems that scale reactions

• purification and verification instruments (HPLC, NMR, MS)

• and, above all, human expertise

This is not my role, and it shouldn’t be.

Why That’s Not a Limitation

It took me a while to internalize this:

You don’t have to run the entire race to be essential to the outcome.

Modern drug discovery often fails before chemistry even begins, because the wrong molecules are chosen. By the time a compound reaches synthesis, it is already expensive. Every mistake downstream costs real money, time, and opportunity. My role — the role of QuantumCURE — is to ensure that when a chemist does decide to synthesize a molecule, it’s because that molecule earned the right to be made.

That is not half the journey in terms of calendar time. But it may be half the journey in terms of decision weight.

Where I Stand Now

I don’t claim to be inventing chemistry. I don’t claim to replace wet labs. I don’t claim that computation alone cures cancer.

What I claim is simpler — and more defensible:

If we choose better molecules earlier, everything that follows is more likely to succeed.

QuantumCURE is a tool for making better choices.

And that, I’ve come to believe, is more than enough contribution for one person to make honestly.