Widening a KRAS G12C series exposure margin from 1.8x to 11x

Covalent KRAS G12C inhibitors occupy the switch-II pocket and engage cysteine 12 irreversibly, so biochemical potency is governed by the rate term k_inact/K_I. A series can therefore look strong on target engagement while failing on the amount of free compound that reaches tumour tissue.

The partner had run four rounds of medicinal chemistry on a single scaffold. Target occupancy in tumour lysate was consistently above 90 percent at the top dose, but the unbound exposure margin over the cellular IC50 sat between 1.5x and 2x across every analogue tested, which left no room for the safety multiple the programme needed at the projected clinical dose.

The margin problem had been read as a potency problem. Each round had pushed k_inact/K_I higher, and each round had produced compounds with the same margin, because the limiting term was free fraction combined with hepatic clearance.

Plasma protein binding across the series was above 99 percent, so small absolute changes in fu moved unbound exposure more than any potency gain achieved in the same round. Human liver microsome half-life was short and varied with a substituent the team had been treating as a potency handle.

Opsin assembled the partner's four rounds into a single evidence set: 214 analogues with measured k_inact/K_I, human and rat microsomal stability, equilibrium dialysis fu in plasma and tumour homogenate, MDCK-MDR1 permeability and efflux ratio, hERG patch clamp, and CYP3A4 time-dependent inhibition.

Compounds were ranked on unbound exposure margin at the projected efficacious dose, which combines every assay in the set. The ranking was then attributed: for each analogue, the model reports which measured property carries the margin and which caps it, so the constraint is located in a named property, with the assay that shows it.

The attribution put the cap on a single position. Metabolic identification confirmed the primary route as CYP3A4-mediated oxidation at the same site the team had been substituting for potency, and 18 analogues were synthesised to break that route while holding the covalent warhead geometry.

The margin gain cost potency. OPS-SM-118 is 12 percent weaker on k_inact/K_I than the compound it replaced, and two of the 18 analogues built to break the CYP3A4 route lost covalent engagement entirely, which the warhead geometry model had not predicted.

Rat exposure improved by less than the human projection implied. Rat microsomal clearance is dominated by a different isoform, so the substitution that resolved the human soft spot bought only a 1.7x rat margin, and the toxicology species had to be reconsidered before the study was scheduled.

A covalent series can be potency-saturated long before it is developable. Once occupancy is above 90 percent at the achievable dose, further gains in k_inact/K_I do not change the margin, and the programme's remaining risk sits entirely in exposure and clearance.

OPS-SM-118 is 12 percent weaker than the round-four compound on the biochemical assay. It carries a margin six times wider, and the hERG and CYP3A4 liabilities that would have surfaced in regulatory toxicology are resolved here.

Microsomal stability was measured in pooled human liver microsomes at 1 uM substrate with NADPH regeneration, sampled at seven time points to 60 minutes. Free fraction was determined by equilibrium dialysis at 37 degrees Celsius over 6 hours, with recovery above 85 percent required for a result to be accepted.

Covalent kinetics were fitted from progress curves at eight inhibitor concentrations. Compounds with incomplete inactivation within 30 minutes were excluded from the k_inact/K_I ranking and carried forward on occupancy alone.