Progress update #2: Proposed validation plan

Previous post: established that NFS1 and RNR are individually validated and identified three open questions: true synergy, delivery of both inhibitors and a conserved targetable site.

Objective:  To convert the hypothesis into a falsifiable, quantitative testing plan with explicit success criteria, controls, and practical requirements  before committing to synthesis and wet-lab work.

Work completed:  I set the design rationale, models, controls and success criteria for 2 tracks with different random modalities.

Evidence (the plan):

• Track A [designed to test for synergy and delivery of both inhibitors]: small molecule prodrug

– Tests and criteria:

– A four-module validation framework was defined.

(i) Synergy quantification module. Both Loewe Additivity (CI = [A]/IC₅₀(A) + [B]/IC₅₀(B); CI < 0.80 = confirmed synergy) and Bliss Independence (SI > 1.0) were specified, with SYNERGY CONFIRMED requiring agreement between both metrics. Crucially, seven normal human safety control cell types (primary hepatocytes, BM-MSCs, cardiomyocytes, dermal fibroblasts, PBMCs, renal PTECs, SH-SY5Y neurons) were pre-registered as internal references: if the Loewe CI of the safety controls is not substantially higher than that of therapeutic targets, any observed 'synergy' would be dismissed as a generic property of combining two cytotoxics rather than a property of the dual-essential context.

(ii) Essentiality and synthetic lethality module. A probability-of-synthetic-lethality framework, P(SL) = w₁·P(SufS essential)·w₂·P(RNR essential)·w₃·P(pathway independence)·w₄·P(no bypass), was specified, with P(SL) ≥ 0.80 required for the SYNTHETIC LETHAL designation thus explicitly distinguishing individual target essentiality from true synthetic lethality.

(iii) Linker activation and resistance module. A four-tier carboxypeptidase (CP) essentiality system (Tier A experimental → Tier D unknown) was defined for the linker, with an automatic downgrade rule: if ≥ 30% of non-viral targets are Tier D, linker redesign is triggered.

(iv) Pharmacokinetic decoupling and selectivity module. A Rodgers–Rowland 15-compartment PBPK model with a Monte Carlo uncertainty analysis (n = 10,000) was specified to test whether the prodrug intracellular ratio stays within the 0.5–2.0 synergy window over the dosing interval, and a Similarity Ensemble Approach screen against ChEMBL 35 (>14,000 targets) with AutoDock Vina docking confirmation was specified for off-target liability with a selectivity of  ≥ 10× margin required.

• Track B [designed to test for a conserved targetable site]: peptide, conserved site on cysteine desulfurases.

– Build a structure panel across bacteria, fungi, parasites, human (cancer) and the host ISC complex(viruses).

– Scan whole surfaces by geometry only, with the catalytic site treated as one candidate among many.

– Test cross-family equivalence, then score sensor compatibility (ODDD 0.4, pH 0.3, protease 0.2, redox 0.1).

– Apply a GO/FLAG/NO-GO gate. GO requires a site present in ≥ 3 categories and sensor-compatible, and an unrelated protein serves as the negative control.

What Changed: the hypothesis is now a testable plan with pre-registered thresholds .Any future claim of 'synergy,' 'selectivity,' or 'essentiality' must survive the specific gates defined here, and the plan itself embeds internal falsification controls (the seven safety cell types) that protect against overinterpretation.

Current interpretation:  Neither track yields a confirmed result by design. Computation prioritises candidates, and bench work must confirm them.

Next step:  With the testing plan established, the framework shifts from design to execution. The next post will document its implementation.