Patient-first
Begin with the cohort.
Separate the diagnosis into endotypes — patient groups with a shared biology — then rank targets by the evidence they matter in each group. The list stays short because every target has to belong to someone.
Approach
Start with the patient
Frontier AI for drug discovery, starting with the patient.
End-to-end AI for choosing better targets, with the patient at the centre of every decision from the first hypothesis to the study.
Explore the platformEndotype
The band a drug is for.
Each band is a patient group. The target is chosen inside the highlighted one.
Our
Mission
A target chosen for an average patient is a bet against the person in the trial. Stratified Precision is an AI drug discovery company built around the opposite order. We keep the patient at the centre: resolve the disease into the groups a medicine can help, select targets that matter inside those groups, and carry the same evidence through to the study.
The patient
comes before
the target.
Approach
Two ways in. One standard for a target.
Some questions start with a cohort. Some start with a gene already on the table. A target is ready when we can say which patients it is for, and whether the risk ahead is harm or a lack of effect.
Target-first
Begin with the gene or protein.
Before chemistry expands, ask who the target is for and how it would fail. Benefit and harm are scored apart. A mechanism with no patient group goes back, not forward.
Platform
Four questions, in the order the patient sets.
End to end means the person who defined the target is the person the study is built to enroll. The platform is that chain, kept intact.
-
01
Patient ground truth
The cohort is the coordinate system.
Clinical course, genetic association, and tissue context stay tied to the people they describe. Later models inherit a human reference, rather than a convenient average.
-
02
Endotypes
A diagnosis is a mixture. A drug needs a group.
We resolve disease into patient groups that share a biology and a reason to respond. Discovery runs against the group. The mixture is what we are trying to leave behind.
-
03
Target selection
Benefit, harm, and specificity, read together.
Inside an endotype, each target is scored on the prospect of effect, the prospect of toxicity, and how specific that signal is to the group. A popular target that only works as an average does not advance.
-
04
Through to the study
The same patients, from hypothesis to enrollment.
The endotype that chose the target also shapes the biomarker and who should be in the study. The patient is present at the last decision because they were present at the first.
The chain
One patient, held through the program.
Patient-first and target-first meet in the same sequence. What changes is the door. What stays is the standard.
-
1
Patient
The cohort the program answers to.
-
2
Endotype
The subgroup with a shared biology.
-
3
Target
Causal inside that subgroup.
-
4
Failure mode
Toxicity, or a lack of effect.
-
5
Study
Enrollment set by the same endotype.
A map of the decision. This is a schematic, not a measured result.
Company
How a result earns the next step.
The group is the objective
Screens, models, and assays answer to a defined patient group. That group is the objective, from the first analysis onward.
Name the failure
Harm and lack of effect ask for different decisions. We keep them apart from the moment a target is scored.
One evidence chain
The endotype, the target, the biomarker, and the enrollment rule should describe the same patients.
Partnerships
Tell us about the patients you are designing for.
Research collaboration, platform questions, and partnership notes reach the same place. We read every one.