From the Hilbeon blog
Reproducible quantum chemistry for drug discovery
Worked examples, DFT walkthroughs and honest method notes — from SMILES to a full quantum portrait, on hardware you already own.
Reproducible to the Last Digit
How do you trust a quantum-chemistry number? Not by comparing programs. Hilbeon earns it two ways: bit-identical results across machines, and agreement with the bench — caffeine's dipole at 3.56 D, aspirin's carbonyls at 1.20/1.22 Å, all reproducible to the last digit.
The Conformer Your Force Field Got Wrong
A force field calls two conformers a tie; plain DFT confidently picks the wrong one by ~1 kcal/mol. Add London dispersion (D4) and the folded, π-stacked conformer actually wins — a 1.5 kcal/mol swing from one physical effect. Why conformer energies need dispersion-aware QM.
Which Tautomer Does Your Assay Actually See?
A tautomer decides which atoms donate and accept hydrogen bonds — the pharmacophore your target reads. Hilbeon optimizes both forms and adds implicit water: the lactam is stabilized by ~4.7 kcal/mol, swinging 2-pyridone to >99.9% and confirming the 4-quinolone antibiotic core.
One Sentence, One Full Quantum Portrait
We told the Hilbeon assistant to “characterize caffeine” and got back a full quantum portrait — geometry, energies, frontier orbitals, dipole — in a single plain-language thread. The computed dipole (3.56 D) lands right on the measured value.
The Cluster You Already Own
A 16-core desktop runs C₆₀ at B3LYP in 215 s and cholesterol DLPNO-MP2 in 247 s. We put its total cost of ownership next to renting the same compute — and the box under your desk wins on price and iteration speed.
A Fullerene Before Coffee: C₆₀ at B3LYP in 215 Seconds
A job that once wanted 15 GB and a cluster now finishes in 215 s on a desktop — no cluster needed. The how: blocked density evaluation, sparse in-core integrals, and point-group symmetry exploited automatically.
Optimizing the Molecule the Photon Sees
Absorption is half the story — photochemistry happens where the excited molecule relaxes to. Hilbeon now ships analytic excited-state gradients: optimize S₁, run excited-state frequencies, predict emission — all validated against finite differences.
Inside Orforglipron: A 113-Atom Drug by Quantum Chemistry
Orforglipron is the first oral GLP-1 pill. Hilbeon builds it from its SMILES and converges a real ab-initio wavefunction — 1006 basis functions, bit-identical on two machines — then reads its physical frontier orbitals and ionization potential.
The Quantum Chemistry Behind the 2025 Nobel: How MOFs Grab CO₂
The 2025 Chemistry Nobel went to metal–organic frameworks for capturing CO₂. Hilbeon computes the amine binding-site reaction and reproduces the dry-vs-humid carbamic-acid / ammonium-carbamate product shift — water vs vacuum, in one run.
Switching a Drug On with Light: Azobenzene by TD-DFT
Photopharmacology turns a drug on with light, and azobenzene is the workhorse hinge. Hilbeon's TD-DFT reads the trans and cis UV-Vis spectra, the stored energy and the singlet–triplet gap in a single excited-state run.
The σ-Hole: How a ‘Negative’ Atom Donates a Bond
A halogen looks negative — yet it donates a bond. Hilbeon maps the σ-hole: fluorine has none (−18.7), chlorine and bromine grow a positive cap (+18.5, +34.0) — the handle behind halogen bonding in drug design.
Why Binding Needs Correlated QM: H-bonds vs Stacking
Hartree-Fock — and dispersion-less force fields — predict that two aromatic rings REPEL (+2.7 kcal/mol). Hilbeon's HF→MP2→CCSD(T) hierarchy shows why drug binding needs correlated quantum chemistry.
How Basic Is Your Amine? Proton Affinity from QM
How protonated is your basic nitrogen at pH 7.4? Hilbeon ranks eight amines by proton affinity and reproduces the experimental order to R² = 0.996 — a trustworthy basicity scale for ADMET.
The Electronics of Drug Rings
Electron-rich or electron-poor? Hilbeon maps ten common drug rings on one frontier-orbital ladder — a ready guide to pi-stacking and metabolic-oxidation liability when you pick a scaffold.
Predicting Substituent Effects: a Hammett Plot from QM
Medicinal chemists use Hammett σ to predict how a substituent shifts acidity and reactivity. Hilbeon reproduces the whole scale from first principles — R² = 0.94 — so you can score a new substituent before you make it.
Carboxylic Acid vs Tetrazole: a Bioisostere
Why does swapping a carboxylic acid for a tetrazole improve permeability? Hilbeon's anion charges show the −1 spread over 4 nitrogens instead of 2 oxygens — the bioisostere's secret.
Ranking Covalent Warheads by Electrophilicity
Covalent drugs live and die by their warhead's reactivity. Hilbeon ranks six Michael-acceptor warheads by their LUMO and electrophilicity index — and shows why acrylamide is the 'Goldilocks' choice.
Seeing Hydrogen Bonds with NCI
Salicylic acid hides its polarity behind an intramolecular hydrogen bond — Hilbeon gives a 0.49 D dipole and an NCI surface you can see. The non-covalent interaction analysis maps exactly where the molecule talks to itself, and why that matters for membrane permeability.
A Computational Portrait of Aspirin
Take a single SMILES string and build a complete quantum picture of aspirin — single-point RHF/B3LYP electronic structure, frontier orbitals, electrostatic potential and non-covalent interactions — in one command. A gentle, reproducible first calculation with Hilbeon.
Predicting Ionization State at pH 7.4
Which form of your molecule actually exists in the body? We walk an absolute pKa through a thermodynamic cycle anchored to a reference acid — and find that at blood pH (7.4) more than 99.99% of aspirin is ionized to its carboxylate.
Where Will the Body Attack Your Molecule?
Frontier orbitals and electrostatic-potential maps point to likely sites of oxidative metabolism. Paracetamol's HOMO sits right on the phenol ring — exactly where cytochrome P450 oxidizes it to the reactive quinone-imine. Insight, not a single bench number.
New posts land here regularly. Subscribe via the RSS feed to follow along.
See your own molecule, start to finish
From a SMILES string to energies, orbitals and properties — in a GUI built for chemists who'd rather do chemistry.