Molecular Mechanisms Lab

Mechanism is a causal connection.

Change a molecular state, make a prediction before the reveal, inspect the evidence boundary, challenge the interpretation and choose the experiment that would discriminate mechanism from artefact.

Module 01 · ALC1

PAR releases autoinhibition.

Primary source: Singh HR et al., Molecular Cell (2017). Single first author. The state diagram is explanatory; publication-backed F2H evidence is a separate evidence layer.

Question

How does PAR convert ALC1 from an autoinhibited molecular state into an active chromatin-remodelling state?

Manipulate · PAR state

Predict before reveal

Challenge · choose the ambiguity to resolve

Decide · next experiment

Engineer · proposed small-molecule mechanism

Explanatory state model · not experimental evidence

Macrodomain⇄ closed coupling ⇄SNF2 remodeler

Without the activating PAR signal, the explanatory model remains biased toward the autoinhibited state.

Integrated causal test · explanatory guidance

PAR absent · no prediction recorded · evidence not yet revealed.

Alternative hypothesisLocalisation change rather than conformational release

Selected evidence testOrthogonal biochemical interaction assay

Proposed small-molecule mechanismEvidence-first / no molecular intervention

Complementary evidence; a more direct test is needed.

The selected experiment will be assessed against the ambiguity you chose.

No small-molecule intervention selected. Resolve the evidence ambiguity first.

Evidence boundary: an interaction change is not automatically a complete mechanism. Localisation, abundance, assay geometry and folding remain alternative explanations until constrained by orthogonal evidence.

Mechanism in motion

ALC1 as a modular chromatin machine.

Modular chromatin engineering: how domain architecture and PAR-dependent control shape ALC1 activity. Companion to the Molecular Cell study.

Mechanism atlas · 9 investigations

From a molecular change to a causal explanation.

Each investigation separates the claim, the measurement and the experiment that could overturn the model. Open only the question you need.

Prediction rule: before opening a card, state the result expected under the preferred model and one result that would favour its strongest alternative.

Define the causal claimWhat exactly is proposed to cause what?

Evidence to inspect: Intervention identity, temporal order, proximal intermediate, downstream phenotype and matched control.
What would falsify it: Verified intervention changes neither the proposed intermediate nor the phenotype.
Primary confound: A correlated state or broad stress response changes all three measurements.
Highest-value next test: Write two competing directed models and choose one intervention with opposite predicted outcomes.

Protect folding and abundanceDid the perturbation alter the intended interface rather than the whole protein?

Evidence to inspect: Expression, stability, solubility, localisation, domain integrity and orthogonal structural or biophysical controls.
What would falsify it: The phenotype remains after abundance and folding are restored independently of the proposed interface.
Primary confound: Deletion or mutation destabilises the protein or moves it to another compartment.
Highest-value next test: Use a minimal interface mutation plus dosage-matched rescue and an orthogonal folding check.

Resolve affinity and kineticsIs the interaction strong, long-lived, cooperative—or merely abundant?

Evidence to inspect: Association and dissociation rates, equilibrium affinity, concentration, stoichiometry, competition and cellular residence time.
What would falsify it: Cellular function is unchanged across a range that materially changes occupancy or residence time.
Primary confound: Avidity, rebinding or local concentration makes an endpoint assay look like tight binding.
Highest-value next test: Measure kinetics and occupancy under matched concentration, then perturb residence without changing abundance.

Test allosteryDoes ligand binding change the conformational ensemble that controls activity?

Evidence to inspect: Ligand occupancy, interdomain contacts, catalytic state, conformational readout and directional rescue.
What would falsify it: Ligand remains bound but the predicted conformational and catalytic transitions do not follow.
Primary confound: Binding and activation are parallel consequences of another modification or partner.
Highest-value next test: For ALC1, separate PAR binding, macrodomain release and ATPase activation with state-specific mutants and inhibitors.

Quantify cooperativityDoes one molecular event alter the probability of the next?

Evidence to inspect: Stoichiometry, Hill behaviour, single-molecule or single-cell distributions, concentration range and alternative binding models.
What would falsify it: A non-cooperative model predicts held-out concentrations equally well.
Primary confound: Heterogeneous abundance or detection thresholds create an artificial steep transition.
Highest-value next test: Fit competing mechanistic models to calibrated concentration series and test a held-out perturbation.

Place mechanism in space and timeWhere and when must the interaction occur to produce function?

Evidence to inspect: Compartment-specific engagement, recruitment kinetics, pulse duration, recovery and spatially resolved downstream response.
What would falsify it: Relocation or timing changes do not affect the response despite disrupting the proposed encounter.
Primary confound: Slow transport or assay integration obscures the true order of events.
Highest-value next test: Use acute recruitment or release with sampling faster than the predicted transition.

Establish necessity, sufficiency and rescueWhich causal leg is actually supported?

Evidence to inspect: Loss, gain, acute perturbation, mechanism-specific rescue, dose response and orthogonal tools.
What would falsify it: The same phenotype appears with unrelated perturbations and cannot be rescued through the proposed link.
Primary confound: Off-target activity, overexpression or adaptation produces a convergent endpoint.
Highest-value next test: Use two orthogonal perturbations and a rescue that restores the specific molecular step.

Prove small-molecule mechanism of actionDoes the compound act through the proposed state and target?

Evidence to inspect: Direct engagement, selectivity, structure–activity relationship, resistant allele, proximal pharmacodynamic marker and phenotype.
What would falsify it: A target-resistant allele or loss of the target fails to change compound response.
Primary confound: Reactive chemistry, aggregation or general DNA-damage toxicity phenocopies target inhibition.
Highest-value next test: Combine engagement, resistant-mutant rescue and proteome-scale selectivity at matched exposure.

Make the programme decisionWhich model deserves the next experiment?

Evidence to inspect: Evidence class, effect size, uncertainty, alternative models, feasibility and pre-defined stopping criteria.
What would falsify it: No plausible result from the proposed experiment would change the action.
Primary confound: The easiest assay is mistaken for the most informative one.
Highest-value next test: Choose the experiment with the greatest separation between model predictions and record outcome-dependent actions.