How the list is built, how its rows are ranked, and what it does not yet know.

Every number on this page is from the edition of 20 October 2026.

What the list is

A list of human protein targets, each paired with the companies most likely to want a new molecule for it. It draws from every human protein that a tested drug acted on, and asks one question of each company at each target: is there a documented reason this company would want something new here? The list is a ranked starting point for scientific review, not a verdict on any target or company.

What removes a row

A row is removed only by one specific, sourced fact from a short, fixed list. Nothing is removed for any other reason; every other row is ranked or shown as undetermined.

The target is not a human protein. Checked before any row is formed.
The same company already has a later approved drug of its own at the target. (276 rows) Route B only. A different drug means a different active substance; a salt, ester or prodrug of the same substance is the same drug. An approval order that is undated or tied proves nothing.
The company dissolved with no successor holding the drug. (18 rows) A dissolution reported only by web research does not count; such a company stays listed and reads “company status not established”.

What “undetermined” means

A row is undetermined when a fact that would decide its route is not yet established. The row stays on the list and names the missing fact; it is never guessed. In this edition 2,746 rows are undetermined, for these reasons:

Whether the company’s drug failed, stopped or is still active is not yet established973

Who held the drug’s approval is not established910

Which protein of a multi-part complex the drug binds is not on record384

The protein the drug is designed for is not established224

The stopped drug is approved and still marketed, so a failure at this target is not proven173

The failure rests only on trials withdrawn before enrolment or with no recorded status82

How rows are ranked

Rows on route A and route B are scored on three questions. Each question gives a value from 0 to 1, where 0.5 is neutral; an unknown fact counts as neutral rather than for or against. The three questions count equally. Within a question, each factor carries the weight shown, rescaled over the factors that apply to that row.

Is the biology proven in humans?

Proof in humans at the target: an approval or a met main endpoint0.50

Balance of efficacy results at the target0.35

How much does this company care?

Furthest phase the company took its own drug0.25

Recent activity at the target or in its disease area0.25

Is there room, and a reason to act now?

Why the company’s own drug stopped0.40

Few other companies active at the target0.30

Route A: a successor drug is active0.15

Route B: the company’s own follow-on0.15

Two discounts can lower a row: when its failure rests only on a drug going quiet rather than on a stopped trial, and when every approval behind a route B row was withdrawn before 1996. Each row shows which factors moved its score and by how much.

US protection ends is shown beside marketed drugs. It is the latest counted date: the last listed US drug-substance patent or exclusivity in the FDA Orange Book; for a biologic, its reference-product or orphan exclusivity in the Purple Book; where neither is listed, 12 years from the first US licence, shown as computed. It is a flag, not a weight: it changes no score and removes no row. It is not the date generics or biosimilars can enter. Use and formulation patents are not counted, and it covers the US only.

Facts read by a language model

Some facts, such as why a trial stopped, were read from free text by a language model. They are labelled “read by a language model” wherever they appear. In this edition no such reading decides a route, removes a row or changes a score; they are shown for context only.

Sources

ClinicalTrials.govEvery registered trial: design, dates, sponsors, stated stop reasons, results

FDA Orange Book and Purple BookUS approvals, holders and their history

European and national registersEU and national approvals and withdrawals

ChEMBL 37Drug identity, targets and measured activity

BindingDB and PubChem BioAssayMeasured activity, including compounds tested inactive

UniProt (via Open Targets 26.06)Where each protein sits in the cell

PharosHow well studied each protein is as a drug target

gnomAD v4.1How tolerant each gene is to loss of function

RCSB PDB, PDBe-KB, AlphaFold DBSolved structures, bound ligands and predicted models

MeSH (in the trial registry)Therapeutic areas of trials, drugs and targets

Company filings and public recordsWho owns whom, renames, acquisitions and dissolutions

Known limits

What this edition is known to miss or simplify. Each is a blind spot of the record, not a judgement; rows it touches stay on the list.

Some approvals are missingA few approved drugs have no approval record in the registers read, so the drug can show as withdrawn or unapproved.

Some subsidiary names are not yet placed in their groupRegional subsidiary names that are not yet linked to their parent leave the drug’s originator unknown. On route B such rows are marked “follow-on unclear” rather than removed or kept on a guess.

Salt forms can count as separate drugsWhere the source records a salt under its own identity, two records of the same drug can appear.

Why a drug stopped is not yet judged by handThe stop cause counts as neutral in the ranking until it is established; language-model readings are shown but not used.

About a fifth of trials have no therapeutic areaTrials whose conditions carry no disease term in the registry’s vocabulary are not placed in an area; area filters do not see them.

The trial funnel counts company-led trialsAcademic-led trials link to drugs and appear on drug and trial pages, but the funnel’s trial stage counts trials led by a company.

Target development level is from a 2024 releaseA newer release reads one step higher for some targets; this does not affect any route or score.

Tissue expression is not yet includedWhere each protein is expressed in the body arrives in a later edition.

Method and limits