CYP450 Isoform Screening Validation | Spirographic AI
Spirographic AI · Validation report

CYP450 Isoform Screening

Deterministic screening validated across 8 cytochrome P450 isoforms with positive reference cases, using a diverse drug panel.

October 19, 2026 · Investigator: Brandy Stoffel, MSN, RN

Validation at a glance

Performance overview

93.9%Per-isoform accuracy · 736 / 784
90.8%Primary CYP in top 3 · 79 / 87
62.1%Primary CYP ranked first · 54 / 87
8CYP isoforms with positive reference cases

What the engine does

The CYP450 screening engine evaluates drug candidates for potential metabolism across 8 validated cytochrome P450 isoforms. It identifies candidate CYP substrate relationships and prioritizes likely primary metabolic pathways, providing a ranked screening result for further investigation.

In this validation, 98 drugs were scored. The per-isoform classification summary covers eight isoforms with positive reference examples; primary CYP Top-3 and Top-1 ranking results reflect the original broader screening panel.

Validation denominator: Of 99 drugs in the validation panel, 98 were scored across 8 CYP isoforms (784 drug–isoform predictions). Vancomycin was the only drug not scored because of corrupted SMILES data.
Detailed metrics

Performance by CYP isoform

Accuracy, sensitivity, specificity, and confusion-matrix counts are shown for the eight isoforms with positive reference examples.

IsoformAccuracySensitivitySpecificityTPTNFPFN
CYP1A292.9%
91/98
8/8 (100.0%)83/90 (92.2%)88370
CYP2B693.9%
92/98
8/9 (88.9%)84/89 (93.9%)88451
CYP2C894.9%
93/98
3/6 (50.0%)90/92 (97.8%)39023
CYP2C990.8%
89/98
10/13 (76.9%)79/85 (92.9%)107963
CYP2C1991.8%
90/98
11/12 (91.7%)79/86 (91.9%)117971
CYP2D692.9%
91/98
17/21 (81.0%)74/77 (96.1%)177434
CYP2E198.0%
96/98
6/8 (75.0%)90/90 (100.0%)69002
CYP3A495.9%
94/98
35/36 (97.2%)59/62 (95.2%)355931

Five additional isoforms were screened but are not included in this eight-isoform classification table because the reference panel contained no positive cases for them.

Validation breadth

Challenging drugs across therapeutic classes

This validation deliberately spans chemically and pharmacologically diverse drugs, rather than relying on one narrow compound family. Examples from the reported drug-level results include:

Oncology & immunomodulation

Paclitaxel, venetoclax, methotrexate, cyclosporine, sirolimus

Neuropsychiatric & analgesic

Duloxetine, risperidone, thioridazine, fluoxetine, methadone, tramadol

Cardiovascular & metabolic

Warfarin, propafenone, atorvastatin, glipizide, repaglinide

Anesthesia & sedatives

Propofol, enflurane, isoflurane, midazolam, triazolam

Antimicrobials & other agents

Isoniazid, efavirenz, acyclovir, gentamicin, lithium

Negative controls

Non-CYP-metabolized drugs were included to challenge false-positive assignments, with both passes and failures disclosed below.

These are illustrative classes and compounds documented in the validation report, not a separately quantified class-stratified benchmark.

Negative controls

Drugs without CYP metabolism

Negative controls were included to test whether the system avoids assigning CYP metabolism to non-substrates. The source validation records seven passes and five failures among the twelve listed controls; vancomycin was excluded from scored isoform predictions.