← Research
Drug Interactions
CYP450 competition + protein coupling — 49 drugs, 1,176 pairs, two layers of danger.
JIM’S OVERSIMPLIFICATION
Two drugs competing for the same parking spot. The stronger binder wins. The weaker one floats around with nowhere to go. That is the first layer. The second layer: the drug is supposed to couple to a sick protein. If another drug blocks the parking spot before the medicine gets there, the treatment fails. Two layers. Both must work.
YOUR LIVER HAS A PARKING PROBLEM
Your liver breaks down drugs using CYP450 enzymes. CYP3A4 alone handles 30 of the 49 most common drugs. That is a parking lot with 30 cars and one spot. When two drugs compete for the same enzyme, one wins and one accumulates to toxic levels.
The worst: sertraline + amiodarone (score: 11.7). Amiodarone blocks THREE enzymes sertraline needs. The deadliest triple: sertraline + warfarin + amiodarone (score: 25.2). A cardiac patient on a blood thinner, an antiarrhythmic, and an antidepressant. This is a real patient profile. All three drugs are commonly prescribed. Often by different doctors who do not talk to each other.
THE SECOND LAYER NOBODY CONNECTS
A breast cancer patient on tamoxifen gets depressed. Doctor prescribes fluoxetine. Fluoxetine blocks CYP2D6 — the exact enzyme tamoxifen needs to activate. The cancer drug stops working. The tumor keeps growing. Two perfectly reasonable drugs that cancel each other out. The drug interaction engine knows about enzyme competition. The protein engine knows about structural coupling. Put them together: the whole patient.
SCALE
49 drugs analyzed
1,176 pairwise interactions
7 CYP enzymes + CYP2C8
RISK BREAKDOWN
98 critical ·
159 high
188 medium ·
731 safe
DRUG CLASSES
39 CYP-metabolized
10 renal / no CYP
5 clinical flags (QT, serotonin, CNS, bleed)
MODEL
Inhibitor + substrate = toxic accumulation
Inducer + substrate = sub-therapeutic
Inhibitor + prodrug = blocked activation
CYP ENZYME SATURATION
Drugs per CYP enzyme
CYP3A430 drugs
CYP2D611 drugs
CYP2C98 drugs
CYP2C196 drugs
CYP1A24 drugs
CYP3A4 metabolizes ~50% of all prescribed drugs.
Every inhibitor of CYP3A4 is a loaded gun aimed at 22 substrates.
TOP 10 MOST DANGEROUS PAIRS
1. Sertraline + Amiodarone — 11.7 (3 CYP pathways blocked)
2. Warfarin + Amiodarone — 9.6 (bleeding risk, both narrow therapeutic index)
3. Sertraline + Fluoxetine — 8.5 (serotonin syndrome)
4. Verapamil + Diltiazem — 8.5 (mutual CYP3A4 blockade)
5. Losartan + Amiodarone — 7.8
6. Hydrocodone + Amiodarone — 7.8 (opioid overdose pathway)
7. Oxycodone + Amiodarone — 7.8
8. Tramadol + Amiodarone — 7.8 (serotonergic + CYP blockade)
9. Sertraline + Phenytoin — 6.5 (antidepressant cleared too fast)
10. Escitalopram + Amiodarone — 6.5 (dual QT prolongation)
Amiodarone appears in 6 of the top 10. It blocks three major enzymes simultaneously, has a 40-55 day half-life, and is narrow therapeutic index.
DEADLIEST TRIPLE
Sertraline + Warfarin + Amiodarone — 25.2
Three drugs. Three CYP enzymes blocked. 10 distinct mechanisms of danger.
Warfarin accumulates → uncontrolled bleeding
Sertraline accumulates through 3 pathways → serotonin toxicity
Amiodarone baseline risk → QT prolongation, cardiac arrest
COMMON COMBOS PEOPLE ACTUALLY TAKE
Fluoxetine + tamoxifen: SSRI blocks CYP2D6. Tamoxifen is a CYP2D6 prodrug. Cancer drug stops working. Published, known, still prescribed together. Score: 5.2.
Omeprazole + clopidogrel: PPI blocks CYP2C19 activation of blood thinner after stent placement. FDA has issued warnings. Still routinely co-prescribed.
Cyclosporine + ketoconazole: CYP3A4 double-load. Nephrotoxicity risk. Score: 5.85.
DRUG × PROTEIN: TWO LAYERS
The drug interaction engine models CYP enzyme competition. The mutation scanner models structural coupling in disease proteins. These are different parts of the same patient.
Cross-predictions by disease:
Alzheimer’s: Treatments use CYP3A4 (donepezil, galantamine). Avoid ketoconazole, erythromycin.
Type 2 diabetes: Metformin is renally cleared (safe). Sulfonylureas use CYP2C9 — avoid amiodarone.
Parkinson’s: Levodopa uses COMT/MAO, not CYP. Lower CYP risk. But entacapone inhibits CYP2C9.
ALS: Riluzole uses CYP1A2. Avoid fluvoxamine (CYP1A2 inhibitor → riluzole toxicity).
COMPUTATION DETAILS
Hardware: Mac Mini M4, $499, 35W
Model: CYP450 enzyme competition + clinical risk multipliers
Data source: CYP interactions from FDA drug labels and PharmGKB
Speed: Milliseconds per pair. 1,176 pairs total.
Software: pip install begump — enzyme load arithmetic, not neural network
HONEST LIMITS
What this model does NOT do:
No dose-response curves | No CYP polymorphism prediction
No P-glycoprotein interactions | No Phase II metabolism
No food interactions (grapefruit = CYP3A4 inhibitor)
No renal/hepatic impairment adjustment
49 drugs is ~10% of commonly prescribed medications
What it IS:
A structural screen. If two drugs share a CYP enzyme, this tells you
which one accumulates and why. First question any pharmacist asks.
Combined with protein coupling: the whole patient in milliseconds.
This is computational research, not medical advice. Drug interactions depend on individual genetics, dosing, renal/hepatic function, and other factors not captured here. Always consult a pharmacist or physician before changing medications.