I Didn’t Build a Docking App. I Built a Quantum Drug Discovery Factory.

I’ve been working my ass off on this for almost a year now. Next November 7th will be the one-year mark, and somewhere past 3,500 hours I stopped counting. Day after day, night after night, I’ve been building something that I hope will actually help people, animals, anyone who ever has to hear the word “cancer” in a doctor’s office.

I’m not obsessed with “beating” the big expensive licensed platforms. That’s not the game I’m playing. Using modern tools, I’ve been able to build a full-blown drug discovery factory with a UI that I’m trying to keep as friendly and intuitive as possible—tooltips on almost every step, guardrails, exports, analyzers—so that this isn’t just a science toy, it’s a usable engine.

And I’m not done. I’m wiring in a GANN AI to pattern-match my best hits, then handing those survivors to SKALA for electron distribution validation. Unlike commercial software that’s often handcuffed by pharma contracts, I mean they p[probably say “you can use our tool, but don’t you dare build your own wet list”. BS!!. I’m doing the opposite.

I’m going to point this engine at PubChem, throw a NVIDIA SPARK (or two chained stations) at it, prep my proteins locally, push the heavy docking to the cloud, and systematically sift through millions of compounds. Why not all of them? I want to build my own wet list, one I can defend, one I own, and maybe, just maybe, find that one needle in the haystack that kicks cancer in the ass.

I’m not a biologist by academia. I’m an engineer who knows how to build complex engines and refuse to quit. Here’s what that looks like when you turn it into a real system:

I Didn’t Build a Docking App. I Built a Quantum Drug Discovery Factory.

Look you all!. Most people build software. I built a machine.

The best way I can describe QuantumCURE Pro is this: it’s the Military Humvee of molecular docking, freaking heavy torque, off-road, and designed to go where the polished city sedans of drug discovery platforms simply can’t.

Everyone else is smoothing UI edges and adding another scoring function. I’m wiring quantum entropy, forensic tracking, VDW geometry, and wet-list management into a single engine whose only job is simple:

Find compounds that deserve to touch a real lab bench.

A Quick Example of What That Actually Means

Here’s the kind of thing this makes possible.

I took a real target, ran the same compound set three ways:

• PRNG-only

• QRNG-seeded

• D-Wave–seeded

In that campaign, the QRNG run surfaced a compound with about 15% better binding in my normalized score window and a novel Bemis–Murcko scaffold that never appeared in the PRNG-only run.

Same target. Same library. The only difference was the entropy source.

That’s the whole point: this isn’t theory. Changing the entropy changes what you actually discover.

Multi-Entropy Docking: Not Just “Random Seeds”

Typical docking systems live and die by PRNG (pseudo-random number generators).I went further and built a multi-entropy spine:

• PRNG – classical baseline

• QRNG – true quantum randomness from hardware

• D-Wave quantum annealing – collapse-driven entropy for seeding

Every simulation knows who fathered the randomness.

That’s not a cute detail. It means I can ask:

• Which entropy source actually discovers the best binders?

• Does quantum seeding consistently find different pockets or scaffolds?

• Is there a real, measurable quantum advantage. Or is it hype?

That’s where the next layer comes in.

Quantum Forensics: Who Really Found the Hit?

I built a Quantum Forensics layer whose only job is to track discovery patterns across entropy sources:

• Which entropy mode found the first plausible binder?

• Which one discovered the most selective candidate?

• Which one uncovered the most novel scaffold?

• How often do quantum-seeded runs beat classical baselines?

It’s like having a black box flight recorder for every docking campaign.

I’m not just running jobs and saving scores. I’m building a history of how each potential drug was discovered—entropy, geometry, potency, and novelty all tied together.

That’s PhD-level research, wrapped into a production system.

VDW Clash Engine: Shape, Space, and Steric Reality

Classical docking scores can lie. Sterics don’t.

So I wired in a full VDW (van der Waals) analysis engine:

• vdw_volume and vdw_surface_area

• Polar surface area (PSA) for permeability clues

• Sphericity & compactness as shape descriptors

• Steric clash detection for physically insane poses

On top of that, I use those VDW metrics to feed drug-likeness and permeability signals into the scoring logic. It’s not just “good binding score = good drug.”

Now the system can say:

• This compound binds tightly but is a steric train wreck.

• This one is compact, clean, permeable, and actually makes sense in 3D.

Docking stops being a one-number game and becomes multi-dimensional reality checking.

From Scores to Dose-Response: Built-In “Prism Lite”

Most docking tools stop at binding affinity.I didn’t.

I built a Dose-Response Analyzer—a mini GraphPad Prism-like module. Right into the platform:

• 4-parameter and 5-parameter logistic curve fitting

• Predicted IC₅₀

• Potency classification (weak / moderate / strong / extremely potent)

• Clean exportable dose–response tables and curves

Why? Because I’m not interested in pretty 3D poses that never leave JSON. I’m interested in compounds you could actually move toward assays and wet labs.

Docking score + IC₅₀ curve + VDW + ADMET == something you can defend in front of a scientist, not just a slide deck.

Wet-List Management: The Endgame, Not an Afterthought

Many platforms are optimized to sell licenses and compute. I’m optimizing for one thing:

A defensible wet list of compounds I can hand to a lab.

QuantumCURE Pro has:

• A Wet-List Manager with security controls

• The ability to promote compounds from raw runs → candidate lists → shortlists → wet list

• Full traceability back to:

  • entropy source,

  • docking conditions,

  • VDW profile,

  • dose–response behavior,

  • scaffold family.

This is not “I ran some docking.” This is “I can prove why this molecule deserves synthesis and testing.”

Bemis–Murcko & Novelty: Not Just Strong, but Different

I also wired in Bemis–Murcko scaffold analysis:

• Group compounds by core scaffold

• See which entropy sources tend to produce novel frameworks

• Score not just potency, but novelty and IP potential

I don’t want a platform that produces the same tired scaffolds everyone else has seen a thousand times.

I want a system that can say:

• This scaffold came from a quantum-seeded run,

• It’s potent, clean on VDW, and structurally distinct from our previous families.

That’s discovery, not just optimization.

Real-Time 3D, Cloud Workers, and Omics Hooks

Under the hood, the engine is wired for real work:

• AutoDock Vina integrated with Cloud Run workers

• Real-time 3D visualization via 3DMol

• An omics integration framework ready to:

  • connect docking hits to pathways,

  • map compounds to targets, mutations, and networks.

This isn’t a toy. It’s infrastructure.

Why I’m so Excited

In under a year, I’ve gone from:

• harvesting quantum entropy,

• to running multi-entropy docking,

• to building forensic tracking, VDW geometry, IC₅₀ analysis, and wet-list management inside a single system.

I didn’t build a “dashboard.” I built a drug discovery factory that:

• knows where its randomness came from,

• understands sterics and shape,

• cares about IC₅₀ and developability,

• and is obsessed with one outcome:

turning simulations into real, testable molecules.

If that sounds like overkill, good. Cancer doesn’t negotiate, and neither should our tools.

And if you’re reading this thinking, “I’ve got the chops, I want in,” then that's good. This is your invitation. I’m opening a QuantumCURE Fellowship track: if you’ve got an M.S. or PhD (or you work at that level in chemistry, physics, CS, or bio), come kick the tires yourself. Log in, stress-test the engine, challenge the assumptions, and help shape what a quantum-enhanced drug discovery factory looks like in the real world—not in a glossy brochure.

I’m not a biologist by academia. I’m an engineer who knows how to build complex engines and refuse to quit. Here’s what that looks like when you turn it into a real system. Ask me question about every metrics and steps... I have an answer for you.

Appendix – Plain-Language Guide to Terms in This Post

This section is for readers who don’t live in quantum / drug discovery land every day. Skim it, jump around, or read straight through.

🔢 Entropy, Randomness & Quantum Stuff

(Like rolling dice with a computer—same starting point = same sequence of 'random' rolls.)

Entropy (in this context) In simulations, like in drug discovery applications “entropy” is basically randomness. You need randomness to explore different possibilities instead of repeating the same thing over and over.

PRNG – Pseudo-Random Number Generator A software-based randomness generator.

• It looks random but is actually produced by a mathematical formula.

• If you start with the same initial number (“seed”), it will produce the same “random” sequence again. Used everywhere in computers, but it’s simulated randomness.

QRNG – Quantum Random Number Generator A hardware device that uses real quantum physics (like photons, beam splitters, etc.) to create randomness.

• It measures truly unpredictable quantum events.

• You can’t rerun the exact same sequence. This is like drawing randomness directly from nature instead of faking it with math.

D-Wave Quantum Annealer A specialized quantum computer made by D-Wave quantum computer company. .

• It’s not a general-purpose quantum computer like you see in theory.

• It solves certain optimization problems by letting a quantum system “settle” into a low-energy state. In My system, it’s used as a source of quantum-flavored randomness and structure, to seed simulations.

Quantum Advantage (in this post) Here it doesn’t mean the formal “industry” quantum advantage definition, but something more practical:

When quantum-based randomness (QRNG, D-Wave) finds better or different drug candidates than classical PRNG would have found on its own.

Quantum Forensics My term for a tracking system that records:

• Which source of randomness (PRNG / QRNG / D-Wave)

• Found which compound

• With what binding strength, selectivity, and novelty

It’s like having a “detective log” of how each potential drug was discovered.

🧪 Drug Discovery & Docking Basics

Drug Discovery (very simplified)

• You have a target (like a protein involved in cancer).

• You test lots of small molecules (compounds) to see which might stick to / interact with that target in a useful way. The goal is to find molecules that might eventually become drugs.

Molecular Docking Computer simulation where you:

• Take a protein (the target)

• Take a small molecule (potential drug)

• Ask the computer: “How might this molecule fit into this protein, and how strongly would it bind?”

Docking gives you an estimated binding affinity (how tightly it might stick).

Binding Affinity (kcal/mol) How strongly a molecule wants to stay attached to a protein.

• Often reported in kcal/mol (kilocalories per mole). Think of it like measuring how hard two magnets stick together.

• More negative numbers (like –8 vs –5) usually mean stronger binding. Think of it as: the more negative, the “stickier” the interaction.

AutoDock Vina A popular open-source docking engine.

• You feed it your protein and your compounds.

• It predicts how they fit together and gives a score (binding affinity).My system wires Vina into a much larger system with quantum entropy and extra analysis.

PubChem A huge public database of chemical compounds run by the U.S. National Institutes of Health (NIH).

• Millions of molecules.

• Free to access. My engine can scan massive subsets of PubChem to look for potential drug candidates.

Wet List This is a short, high-confidence list of compounds that have survived all the computational filtering and are:

• Worth sending to a real lab.

• Ready for wet experiments (actual test tubes, cells, animals, etc.).

Everything before the wet list is “digital.” After that, you’re in real-world biology.

🧬 Metrics, Formulas & Curves

IC₅₀ (Half-Maximal Inhibitory Concentration) A standard measure in pharmacology. to check how electrons (tiny charged particles) are arranged in the molecule.

• Imagine you increase the dose of a drug.

• At some concentration, the drug reaches 50% of its maximum effect. That concentration is the IC₅₀.Lower IC₅₀ → more potent drug (you need less of it to see an effect). Effect may be not a suitable word.. so think of it as "at some concentration, the drug blocks 50% of the target's activity."

Dose–Response Curve A graph showing:

• X-axis: dose (concentration of the compound)

• Y-axis: response (effect: inhibition, activity, etc.)

It often has an S-shaped form. My engine fits this curve and extracts IC₅₀ and potency labels (e.g., weak, moderate, strong, extremely potent).

4PL / 5PL Logistic Curve Fitting These are mathematical models used to fit the dose–response curve.

• 4PL (4-parameter logistic):Models the lower limit, upper limit, slope, and midpoint (IC₅₀).

• 5PL (5-parameter logistic):Same as 4PL but adds asymmetry, allowing the S-shape to be skewed if the data isn’t perfectly symmetric.

You don’t need the math; just know: this is industry-standard curve fitting used to interpret dose–response experiments.

ADMET Stands for:

• Absorption

• Distribution

• Metabolism

• Excretion

• Toxicity

In plain English:

“If this molecule goes into a body, how is it absorbed, where does it go, how is it broken down, how is it removed, and how dangerous is it?”

ADMET is about whether an otherwise good binder can survive and behave in a real organism.

⚙️ Geometry, VDW & Scaffolds

VDW – Van der Waals Weak attractive or repulsive forces between atoms, based on how close they are. VDW is named after a Dutch scientist—basically the physics of atomic 'personal space!.

QuantumCURE VDW engine calculates:

• vdw_volume – How much 3D space the molecule occupies.

• vdw_surface_area – How much surface it presents to the world.

• VDW sphericity / compactness – How “ball-like” or “spread out” it is.

• Steric clashes – Whether atoms are unrealistically jammed into each other in a way that would be physically impossible.

These checks help filter out physically nonsense docking poses and bad shapes.

PSA – Polar Surface Area A measure related to how polar (charged) parts of a molecule are distributed on its surface.

• Often used as a hint for permeability:

  • Low PSA → more likely to cross membranes (e.g., get into cells, cross the gut wall, sometimes the blood–brain barrier).

QuantumCURE Pro uses PSA as part of drug-likeness prediction.

Bemis–Murcko Scaffold A way of boiling a molecule down to its core framework:

• Remove side chains and decorations.

• Look at the main ring systems and linkers.

This helps group molecules into families and check for novelty:

• “Do we keep rediscovering the same old scaffolds?”

• “Did quantum runs generate a new core architecture we haven’t seen before?”

Scaffold Novelty Simply: “Is this backbone structure different from what we’ve already tried or what’s common in known drugs? ”High novelty = more interesting for IP (patents) and scientific exploration, as long as ADMET isn’t terrible.

🧠 AI, Omics & Extra Modules

GANN AI Refers to using Generative Adversarial Neural Networks or similar modern AI models to:

• Look at lots of candidate compounds.

• Find patterns in what looks promising.

• Suggest or prioritize molecules for further testing.

Think of it as an AI filter on top of everything else.

SKALA (Electron Distribution Validation)Here, SKALA refers to tools / methods designed to look deeper into electron distribution and quantum behavior of promising molecules.

The idea is:

• First, filter millions of compounds using docking & entropy.

• Then, for a smaller, high-confidence list, use more advanced physics-level analysis to double-check that the electron behavior makes sense for binding and selectivity.

Omics Integration “Omics” covers big biological data layers like:

• Genomics – genes

• Transcriptomics – RNA

• Proteomics – proteins

• Metabolomics – small molecules

An omics integration framework means the platform is being wired so that:

• Docking hits can be connected to real biological pathways, gene expression, and disease mechanisms.

So it’s not just “this molecule binds this protein,” but:

• “Here’s how this might ripple across a whole biological system.”

🖥️ Infrastructure & Tools

3DMol A browser-based 3D viewer for molecules and proteins.

• Lets you rotate, zoom, and inspect structures in real time. My system (engine) uses it so users can visually inspect poses and interactions.

Cloud Run Workers Refers to Google Cloud Run, a hosting system for running containerized apps on demand.

• My system (engine) uses “workers” to spin up docking jobs in the cloud.

• This makes it scalable instead of being stuck on one local machine.

NVIDIA SPARK (Workstation) In this context, SPARK refers to a powerful NVIDIA workstation / GPU box that can massively speed up calculations.

• Used locally to prepare proteins and pre-processing tasks.

• Heavy docking can then run on the cloud.

Think of it as quantumCURE's local muscle, with the cloud as the scaling engine.

👥 Programs & People

QuantumCURE Fellowship Mansour’s (the founder) planned program to invite:

• M.S. / PhD-level scientists, engineers, and researchers to kick the tires of the platform, run their own tests, stress it, question it, and help shape its evolution.

It’s a way of turning this from “one guy’s engine” into a community-influenced research platform.

#QuantumCURE#QuantumLaso#DrugDiscovery#QuantumComputing#AIInDrugDiscovery#ComputationalChemistry#MolecularDocking#CancerResearch#TranslationalResearch#BiotechInnovation

=================================================================================

Quantum + entropy :#QuantumEntropy#QuantumAnnealing#QRNG#QuantumForensics#EntropyAwareDiscovery#QuantumAdvantage

Fellowship / talent #FellowshipProgram#ResearchFellowship#PhDLife#PostdocLife#CitizenScientist#OpenScience

:#SaaS #DeepTech #BioAI#QML #HPC #CloudComputing