A cited target safety assessment for any new target, scored by organ system and signed by a toxicologist
A cited safety assessment for any new target in under two days, with every risk scored and a plan to de-risk it.
See one case, screen by screen ↓A project team wants to know whether a new target is safe enough to work on, and the portfolio committee wants the answer by its pack date. The discovery toxicologist has to cover every line of evidence: what happens to people born with less of the target, what knockout and conditional mice show, where the target is expressed in people and in the four toxicology species, what labels and adverse event reports say about drugs on the same pathway, what else the molecule could hit, and what your own studies already showed.
Most of that time is searching and reading, not judging. Each finding has to be pulled out with its source, sorted by organ system, scored and written up in the house template, with a monitoring and de-risking plan behind it. By hand that is about two weeks per target, and a liability that rests on one weak source can look as solid as one that four evidence lines agree on.
Estimated split for one target, in working days, by hand and with the solution.
Nine agents plan, gather every line of evidence in parallel, score each risk, draft and challenge the report; a toxicologist reviews and signs.
SRK9 is a kinase target for rheumatoid arthritis, with an oral small molecule, NWB-5521, heading for the October 22 portfolio committee. A new request, NLRP15 for gout, arrives the same morning. Here is the day, screen by screen, in the working solution.
Lena opens the solution: 4 assessments wait for her review and the committee packs close October 15. The body map shows SRK9’s on-target risk by organ system: 10 liabilities from 16 key sources, with serious infections and embryo-fetal lethality high, and brain, eyes, lungs, kidney and bone with no signal found.
“4 assessments wait for your review · portfolio committee packs close Oct 15”
Maya’s request for NLRP15 in gout is waiting. Assessing it takes the target, the disease and the modality, here an oral small molecule. Six evidence agents start at once, and the draft lands with the toxicologist for review.
Human genetics reads 1,206 carriers, animal models 14 phenotypes, expression 42 tissues and 9 cell types, same-mechanism drugs 3 labels and 2,940 adverse event reports, pathway and off-target 38 pathway partners, and our own studies finds none yet, with the first rat study in January. Findings land as they are found, each tagged to an organ system and cited.
“Gout risk 31% lower among people with one broken copy — supports the target”
Back on SRK9, the liability reads Likely × Serious, high risk, agent confidence 93%. The evidence ladder shows 4 of 6 lines supporting it: children with no SRK9 have recurrent pyogenic infections, myeloid knockout mice cannot clear Listeria, the target is highest in neutrophils and monocytes, and a same-mechanism drug carries serious infections at 3.1 per 100 patient-years.
“People with one working copy are healthy, so partial inhibition may be tolerated.”
Each citation opens the source at the passage: here, a 2022 case series of six children from three families with homozygous loss-of-function variants. The lines the liability rests on are highlighted, and the screen lists every liability that cites the source.
“Neutrophil killing of Staphylococcus aureus was reduced to 35% of control.”
Malignancy (class warning) rests on one source, the class boxed warning on a same-mechanism label, and human genetics argues against it. The critic says so. It also had the knockout heart defect weighed under embryo-fetal risk, because it is developmental, not an adult heart-rhythm finding. Both notes stay open for Lena’s judgement.
“Only one source supports this liability. The critic suggests keeping probability at Unlikely unless a mechanism is found.”
Lena moves malignancy from Unlikely to Remote on the 5 × 5 matrix. A re-score needs a reason, and the reason goes on the review record: a single class-label source, no excess among carriers, and no difference in the 5-year extension. A programme-stopping impact is never scored low, so it stays a medium risk.
Seven studies are suggested, each linked to the liability it answers; four are in the plan, the longest 8 weeks, about $198k. They include a bone-marrow colony assay with human against rat cells, a stem-cell cardiomyocyte assay with an SRK8 counter-screen, a neutrophil killing assay in human blood and a rat embryo-fetal dose-range finder before Phase 2.
Summary and recommendation, target and pathway, expression, human genetics, animal models, same-mechanism drugs, liability register, monitoring and de-risking plan, and species relevance and study design. Every statement carries a numbered citation, and the draft exports to Word or goes to the IND nonclinical overview as a draft section, references kept.
The solution will not let Lena sign until she has reviewed both high risks. She keeps the recommendation, “Go with a de-risking plan”, and signs with her password. The meaning of the signature is recorded with it: approved by toxicologist. Because two high risks remain, Dr. Omar Haddad, head of discovery safety, co-signs, and the report goes into the committee pack.
“By signing I confirm I have reviewed the evidence and approve this target safety assessment.”
The same assessment, seen by the people who request, assess, sign and act on it — what their week looked like, and what it looks like now.
Nine agents plan, gather every line of evidence in parallel, score each risk, draft and challenge the report; a toxicologist reviews and signs.
Turns a request into an evidence plan: resolves the target and its paralogs, sets the indication and modality, and starts six evidence agents in parallel.
Finds what happens to people with less of the target: loss-of-function counts and constraint, biobank carrier scans, Mendelian disease and case reports.
Collects knockout, knock-in and conditional phenotypes, including embryonic lethality.
Summarises where the target is, by organ system, in people and in mouse, rat, dog and monkey, and flags species where the pattern differs.
Reads labels, boxed warnings, public adverse event reports and trial results for drugs on the same target or pathway, and extracts every safety finding with its rate and section.
Checks paralogs, selectivity panels, hERG and secondary pharmacology, and what your own study archive already showed.
Merges findings into liabilities by organ system, scores probability × impact and monitorability, and suggests de-risking studies.
Writes each section of the assessment in the house template from the register and the sources, with numbered references.
Challenges single-source liabilities, species relevance and misplaced developmental findings; resolves what it can and leaves the rest for the toxicologist.
Re-scores with a reason and signs. The agents propose; a named person decides.
The toxicologist can ask about any liability, plan or portfolio question, or tell it what to change. Answers cite their sources.
Why is SRK9 a high infection risk?
It is Likely × Serious because four evidence lines agree: no SRK9 means recurrent pyogenic infection in 3 families; a myeloid knockout means a 40-fold bacterial burden; expression is highest in neutrophils and monocytes; a same-mechanism drug has serious infections at 3.1 per 100 patient-years. What argues for a window: 418 people with one broken copy are healthy.
Is the embryo risk a programme stopper?
No — it is high but manageable. Knockout mice die at E11.5–E12.5 with heart defects and no person with two broken copies is known among 141,456 exomes, so the embryo-fetal studies will very likely show findings. Trials use highly effective contraception and the rat embryo-fetal dose-range finder runs before Phase 2 ($120k, 8 weeks).
Add the antibody-response study to the SRK9 plan
Added. The T-cell dependent antibody response study runs inside the 4-week rat study — 3 weeks, $38k, owner Dr. Lena Ortiz. It answers Serious infections, one of the 2 high risks. The plan now holds 5 of 7 studies ($236k) and the change is on the review record.
How much time does this save?
Over the last 30 days the median from request to a signed assessment was 1.6 days, against about 2 weeks when a toxicologist reads and writes it by hand — a typical target, not a guarantee. Agent time is about 10–20 minutes per target; the rest is you deciding.
13 screens from the working solution, on its sample data. Pick one to see it large.
The report template, how risks are scored, which sources the agents read and who signs are all settings.
| Setting | Default | Choose from |
|---|---|---|
| High risk means probability × impact of | 12 or more | 10 · 12 · 15 or more |
| A high risk needs a second signature | On · Dr. Omar Haddad | On or off · the named second signer |
| Optional report sections | Species relevance and study design | Species relevance · Competitor safety comparison |
| Accept low risks in one step | On | On · Off |
| Critic challenges every single-source liability | On | On · Off |
| Re-check signed assessments for new evidence | Weekly | Weekly · Every two weeks · Monthly |
| Preprints | Off | On, labelled “not peer reviewed” wherever cited · Off |
| Who signs, by therapeutic area | Immunology: Dr. Lena Ortiz | Dr. Lena Ortiz or Dr. Omar Haddad, set per area |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
“demo” = seen in the working solution, on its sample portfolio · “target” = the design goal, measured in the live solution · “estimated” = our estimate · “typical” = published figures (published target-safety-assessment methods and services — weeks per target by hand; toxicity ≈ 30–40% of preclinical failures). People, targets and molecules named on this page, and Northwind Bio, are fictional characters and data in the working solution.
A review of the safety risks of acting on a drug target, done before a programme commits to it: on-target liabilities, knockout phenotypes, human genetics, expression, class effects of drugs with the same mechanism and off-target risks. The Target Safety Assessment Agent gathers that evidence with agents, scores each liability and drafts the TSA for a toxicologist to review and sign.
The risk scorer groups findings into liabilities by organ system and scores each on probability against impact on a 5 × 5 matrix, with monitorability and agent confidence. High risk means a product of 12 or more by default, and a programme-stopping impact is never low. The toxicologist accepts, re-scores with a reason or marks a liability not relevant.
Licensed literature, mouse phenotype and public exome databases, population biobank scans, tissue expression atlases for people and four toxicology species, drug labels, public adverse event reports, trial registries and your own toxicology archive. Preprints can be switched on and are labelled “not peer reviewed” wherever cited.
Yes. Each liability can carry suggested studies, such as a bone-marrow colony assay with human against rat cells or a rat embryo-fetal dose-range finder, with weeks, cost and owner. The toxicologist chooses which go into the plan and sends it to the project team.
Yes. Agents only propose scores; a toxicologist decides on each liability, and the critic’s challenges are shown, never applied silently. High risks must be reviewed before signing, the signature records its meaning, and a high risk brings a second signature from the head of discovery safety.
Evidence watch re-checks signed assessments for new labels, adverse event reports and knockout papers, weekly by default. When new evidence changes a score, the assessment is re-opened and its owner gets a Needs-you item.
Yes. The assessment is drafted in your house template; the core sections are always included and optional ones, such as species relevance and study design or a competitor safety comparison, are settings. It exports to Word or PDF with citations kept as numbered references.
The Agentic Solution Engine builds and deploys it from your requirements — your report template, scoring rules, sources and a few past assessments — and it goes live once every quality gate has passed. We will walk you through it on one of your own targets first.
Ask the literature in plain words and get a sourced answer in seconds.
1-min filmPipeline WatchEvery rival trial, readout and approval lands in a sourced brief for each programme, every Monday.
1-min filmPreclinical ReportsRaw study data becomes a draft report in hours, with every number traced to its source.
1-min filmWe’ll run a target safety assessment on one of your own targets, in your house template.