The Clause No Emergency Reaches

I.The First Known Thing — the rights that may be suspended, and the ones that may not

public international law · derogation in time of emergency

A human rights treaty is a set of promises a state makes about how it will treat people, and the drafters of the modern ones knew that states under threat would break them. So they wrote the breaking into the text. Article 4 of the International Covenant on Civil and Political Rights permits a state, in time of public emergency threatening the life of the nation and officially proclaimed, to take measures derogating from its obligations under the Covenant — but only to the extent strictly required by the exigencies of the situation, only consistently with its other obligations under international law, and only without discrimination on grounds of race, colour, sex, language, religion or social origin. Article 15 of the European Convention does the same in almost the same words. Derogation must be notified to the other parties through the Secretary-General, with the provisions suspended and the reasons given, and notified again when it ends. Nearly the whole of the Covenant is derogable in this way. Liberty of the person, freedom of assembly and association, freedom of movement, privacy, expression, the public character of a trial: under a sufficiently grave emergency, every one of these may be curtailed, and the doctrine that governs the curtailment is proportionality. The question asked of each measure is whether it is strictly required, and the answer is a matter of degree.

and then a list that is not a ranking

Article 4 then does something the rest of the instrument does not prepare you for. Its second paragraph names a handful of provisions from which no derogation may be made under any circumstances: article 6, the right to life; article 7, the prohibition of torture and of cruel, inhuman or degrading treatment, and of medical or scientific experimentation without consent; article 8, paragraphs 1 and 2, the prohibition of slavery, the slave trade and servitude; article 11, the prohibition of imprisonment merely on the ground of inability to fulfil a contractual obligation; article 15, the principle of legality in criminal law, that no one be convicted for an act that was not an offence when committed; article 16, the right to recognition everywhere as a person before the law; and article 18, freedom of thought, conscience and religion. Read the list twice. It is not a ranking of importance. Debtors' prison stands in it beside torture, and the right to a fair hearing — which most people, asked to rank the Covenant, would place very high — does not stand in it at all. Whatever principle assembled this list, it was not severity, and it was not the order in which the rights would be missed.

what the list does track

The Human Rights Committee, setting out the doctrine in its General Comment on article 4, gives two reasons for the enumeration, and they are different kinds of reason. Some of the provisions are there in recognition of the peremptory character of the underlying norm — the prohibitions of torture and of slavery bind whether or not a state has signed anything. Others are there, the Committee says, because it can never become necessary to derogate from them during a state of emergency: suspending article 11 or article 18 would buy a government nothing it could use against the threat, so no proportionality test could ever be passed and the question need not be asked. Around this sits a body of doctrine built on a third idea, which the Committee applies constantly without writing it into article 4. Its general comment on the nature of states' obligations forbids removing a person to a country where there are substantial grounds for believing a real risk of irreparable harm; its interim measures under the Optional Protocol exist to hold a situation still where harm would otherwise become irreparable before the Committee could rule. And article 4 itself is built on the assumption that a derogation is a loan: exceptional, temporary, notified, and aimed at the restoration of a state of normalcy in which the Covenant can be fully honoured again. The loan can only be taken against a right that can be handed back. A suspended assembly is reconvened; a curfew lifts; a delayed trial is held. An execution is not reversed by the end of the emergency, and neither is a person tortured, enslaved, or erased from legal personhood. The working distinction in the law of emergencies is between the promises whose breach the return to normalcy repairs and the promises whose breach it does not — and the law carries that distinction categorically, as a place the proportionality test may not enter, rather than as a very large thumb on its scale.

II.The Second Known Thing — the scorecard

drug discovery · lead optimisation and multi-parameter trade-offs

Walk to a hall where nothing is proclaimed. A series of molecules has shown activity against a target, and the work of the next two or three years is to turn one of them into something that can be given to a person. The molecule must be potent, and selective, and soluble enough to dose, and permeable enough to absorb, and stable enough against the liver's enzymes to last, and free enough of plasma protein to act, and quiet at the cardiac potassium channel, and quiet at the cytochromes it might inhibit, and clean in the assays for genetic damage, and cheap enough to make. Each of these has a threshold that the organisation has set for itself. None of them is independent of the others: they are all consequences of the same atoms, and lipophilicity in particular moves half of them at once, upward here and downward there. A chemist improving one number is usually spending another, and the discipline of lead optimisation is the management of that account.

how the field adds the properties together

The field's answer, worked out over twenty-five years, is to score. Ligand efficiency and lipophilic efficiency normalise potency by what was spent to get it. Beyond those, the standard apparatus is the desirability function: each property is mapped onto a scale from zero to one by a curve expressing how good that value is, and the mapped values are combined into a single figure. The quantitative estimate of drug-likeness published by Bickerton and colleagues in 2012 takes eight molecular properties and combines their desirabilities as a weighted geometric mean. Pfizer's central nervous system scorecard, published by Wager and colleagues in 2010, maps six physicochemical properties each onto a zero-to-one scale and sums them to a figure out of six. Commercial platforms build the same object from predicted rather than measured values and propagate the uncertainty through it. The output in every case is one number per compound, and the compounds are ranked by it.

and the argument the field had about filters

The scorecard replaced something harder. The rules of the late nineteen-nineties — Lipinski's and its successors — were cut-offs: a molecule with more than five hydrogen-bond donors, or a molecular weight above five hundred, was out. The case against cut-offs was made repeatedly and has largely been won. A filter applied to a predicted property discards compounds on the strength of a number with real error bars around it; a compound just the wrong side of a line is treated as categorically different from one just the right side; and good drugs exist beyond every line ever drawn. A score keeps the information a filter throws away, handles uncertainty properly, and lets a genuine excellence in one property argue for a compound that is ordinary in another. So the modern scorecard prices everything. What it cannot express is a property that must not be priced, because a weight is an exchange rate and an exchange rate always has an amount on the other side: for any weight given to the cardiac channel, there is a potency that pays for it. The field does hold some things outside the scorecard — structural alerts for mutagenicity, a hard stop on a positive genotoxicity finding, a safety margin that is simply required — but which properties get that treatment is a matter of local convention, inherited from the last programme that was killed by one. It is not derived from anything, it varies between organisations, and the logic that won the argument against filters argues quietly against its existing at all.

III.The Bridge

no scorecard cites article 4

The halls share nothing. The literature of emergency derogation and the literature of multi-parameter optimisation in medicinal chemistry have no common citations and no review spans them; the nearest either comes to the other is the coincidence that both use the word proportionality, and they use it for the same idea, which is precisely the idea that both halls have well in hand. What neither hall holds in common is the other thing. Both are constraint systems that must be relaxed under pressure — a state under threat cannot honour every promise, a series under development cannot meet every threshold — and both must therefore decide which constraints may be traded away and on what terms. One of them has had two centuries of catastrophe in which to learn that this decision has two kinds of answer and that only one of them can be written as a weight.

X: a weight is a price, and some breaches cannot be bought back

Call it non-derogability. Where a system of constraints must be relaxed under pressure, the constraints that have to be held exempt from the relaxation are fixed not by their importance but by whether their breach forecloses the later return to normal; and that exemption cannot be expressed as a weight, however large, because a weight is a price and every price can be met. From which follows the half that the optimisation hall has not ported: a scheme that prices every property has not decided that nothing is unrecoverable — it has been unable to say so, and the distinction between a deficiency that later work can repair and one that ends the programme has to be carried categorically or it is not carried at all. This also reframes the argument the field settled. The case against hard filters was an argument from uncertainty: we do not know the value well enough to cut on it. The case for a non-derogable core is an argument from reversibility, and reversibility is orthogonal to uncertainty — a badly measured property may still be one whose failure a formulation can rescue, and a well measured one may not be. The two arguments have been answering past each other, and the field, having won the one it was having, never had the other. The question the derogation hall makes askable is not how confident we are in a number, but whether a programme that advances with a bad number on that axis can still be saved without starting again from a new molecule.

IV.The Question That Becomes Possible

neither discipline would ask this alone

Phrased this way the missing quantity is obvious and has never, so far as any published scorecard shows, been estimated. Every property on a scorecard has a weight. None of them has a recoverability: a number saying how often a programme that went forward with that property below its own threshold was afterwards rescued with the same molecule — by a salt or an amorphous dispersion, a change of route, a smaller dose more often, a narrower population, a different formulation — rather than being sent back for a new one or stopped. The record for computing it exists inside companies as candidate-selection profiles set against what the programmes went on to do, and the nearest published object of that shape is the retrospective analysis by Waring and colleagues of the candidate drugs of four major companies, which set property profiles against the reasons those candidates failed.

the non-derogability hypothesis

Does each optimisation parameter carry a recoverability — the rate at which programmes that advanced below their own threshold on that parameter were afterwards rescued without a new molecule — and does a scorecard that gates the low-recoverability parameters categorically while weighting only the rest predict candidate attrition better than one that weights them all?

why it would matter

Because recoverability and weight are different orderings, and the places they come apart are the places the scorecard is currently silent. Solubility is weighted heavily and is often recoverable; the formulation sciences exist largely to recover it, and a programme that reaches development with a poorly soluble molecule frequently reaches the clinic with the same molecule in a different form. Clearance is partly recoverable, by dose and by schedule, up to the point where the dose required brings the safety margin down with it. A cardiac channel liability or a genotoxic alert is much less recoverable: no formulation removes it, and what the programme does next is chemistry, not development. If the recoverabilities are as different as that reasoning suggests, then two candidates with the same aggregate score can carry entirely different amounts of unrepairable deficiency, and the score cannot see the difference — which would explain a pattern the field knows anecdotally, where a scorecard-excellent candidate dies on one axis that was never going to be fixable and nobody can say afterwards why the score did not catch it. And the derogation hall supplies the prior that cuts against the candidate, which is why it is worth stating plainly: the reason a short list of go/no-go criteria already sits outside every company's scorecard may be that experienced people have arrived at the derogation list empirically and hold it as convention. If so, the proposal is not a new measure but a derivation of one the field already has, and the work is to check whether the conventional list is the one the record supports.

V.How It Could Be Proven — or Broken

a candidate must be breakable, or it is merely poetry

An intercognix candidate earns nothing until it names its own test. This one is retrospective, it needs records that companies hold and mostly do not publish, and the shape of the analysis is close enough to published attrition work that the obstacle is access rather than method.

the honest status

Note the narrowness. The claim is about small-molecule candidate-selection profiles and the parameters that appear on a scorecard. It proposes no threshold values, no new property to measure, and no change to how any property is predicted. It does not say that categorical gates are generally better than scores — the argument that won against the rules of the nineteen-nineties still stands where it applies. It says only that the choice between gating a parameter and weighting it is currently made by convention, that the derogation hall names the axis on which that choice should be made, that the axis is recoverability rather than importance, and that the two orderings can be compared against a record which exists. And a caution about what the analogy does not carry: a non-derogable right is a moral commitment whose grounds are not merely prudential, and nothing in this specimen suggests that the law's list is really a calculation about reversibility in disguise. What is borrowed is structural — that a system which must relax under pressure discovers it needs a short list of exemptions, that the list is not a ranking of what matters most, and that no weight can be made to do the list's work. Until a portfolio's parameters have been given recoverabilities and the two-tier rule has been tested out of sample, this remains exactly what its stamp says: a candidate. A structural echo between two halls of the library, waiting for someone with the tools of both to walk the corridor.