Preclinical study report automation that links every number to its raw data cell
A draft GLP report in hours, every number traced to raw data.
See one case, screen by screen ↓When the in-life phase and necropsy are over, the study director and the contributing scientists turn a raw data package — body weights, food consumption, clinical signs, clinical pathology, organ weights, macroscopic and microscopic pathology, toxicokinetics — into a report in the house template. A repeat-dose report has 9 sections, 10 to 20 tables and 400 to 4,000 numbers, and most of the one to two weeks goes on re-keying those numbers from the export and cross-checking them.
Quality Assurance then audits every report number against the raw data, and every transcription slip becomes a finding. The quieter risk is a missing value: a weighing that never reached the export, filled from the week before or from a sister group, reads as data. In a GLP report that is a data-integrity issue, not a typo — and the study director signs for every number.
Estimated split from data received to a signed draft, by hand and with the solution. The by-hand total matches the team's 11.2-day baseline on the solution's dashboard; the total with agents is the 1.4-day target.
Six agents read, build, draft and check every report; a value the raw data lacks stays blank; the study director e-signs.
VLT-TOX-028: Veltrimab, a humanised antibody, given subcutaneously once a week to Sprague Dawley rats at 0, 10, 30 and 100 mg/kg/week for 28 days, with a 14-day recovery period — 80 main-study and 20 recovery animals. The data package arrived on 05 Oct 2026 and the report is due on 16 Oct. Here is what happens next, screen by screen, in the working solution.
Overnight the agents read study plan v2.0 with amendment 1, the 13 raw data tables, repeat-dose template v5.2 and Sam Patel's signed toxicokinetics report, and drafted all 9 sections. Lena opens a split screen: the report in the house template beside the export, read-only and locked. The header reads 156 traced, 1 blank, 2 gaps, 3 of 9 sections approved.
“VLT-TOX-028 drafted overnight — 9 sections, 1 value left blank, 2 gaps.”
In males at 100 mg/kg/week, body weight gain over Days 1–29 was 142.7 g against 156.8 g in controls (−9.0%; p<0.05). Lena points at the −9.0% and both cells it comes from light in BW_Males — E7 and B7 — with the calculation and every section that uses the cell: 1, 4.2 and 5.
“(142.7 − 156.8) ÷ 156.8 = −9.0% · cells E7 and B7.”
Cell E5 of FC_Males, Week 4 food consumption for Group 4 males, is empty in the export of 05 Oct 2026. The study log records the weighing on 12 Sep by Ana Kowalski; that day's balance file was never transferred. The drafter wrote no number and no 4-week mean — the sentence reads “not in raw data”.
“The agent left the report blank — no estimate, no carry-forward from the week before.”
The gap finder lists the blank, which blocks signature, and a second blocking gap: study plan §9.4 requires microscopic examination of Groups 1R and 4R at the recovery necropsy, and MI_Recovery.xlsx is not in the package. A minor item asks Lena to confirm one excluded value — a Week 3 weighing for cage M-4-07, recorded as spilled — and confirming it adds a note on the exclusion to the food consumption results.
Study plan §9.4: “…examined microscopically for all animals in Groups 1 and 4 at the terminal and recovery necropsies.”
Lena clicks “Ask Ana Kowalski”. The corrected export, VLT-TOX-028_FC_v2.xlsx, comes back signed and versioned — the first export stays in the record. Week 4 is 25.9 g/animal/day; the raw data reader checks it against the plan, the section drafter redrafts 4.2 with the 4-week mean of 25.3 against 26.6 g/animal/day in controls (−4.9%), and the consistency checker finds 0 mismatches.
“Blank filled from VLT-TOX-028_FC_v2.xlsx — section 4.2 redrafted.”
Lena requests the recovery data from Dr. Omar Haddad. MI_Recovery.xlsx arrives for 20 animals, signed by him. The drafter rewrites 4.5 Histopathology — after 14 days without treatment, no hepatocellular hypertrophy at 100 mg/kg/week (0/5 males, 0/5 females) and a minimal injection-site infiltrate in 1/5 males — and updates the Summary and the Conclusion with reversibility and the NOAEL.
“Gap resolved — recovery findings added to sections 4.5, 1 and 5.”
Findings at 100 mg/kg/week — lower male weight gain, ALT 53 against 38 U/L and minimal to mild centrilobular hepatocellular hypertrophy in 4 of 10 males — had no degenerative correlate. Before the recovery data, the draft proposed a NOAEL of 100 mg/kg/week, the highest dose tested, “subject to the recovery data”. With the data in, the conclusion states reversibility and that NOAEL — still the study director's call.
The section drafter never decides adversity or the NOAEL on its own — it proposes them and marks them for the study director.
The signature screen shows what she is signing: VLT-TOX-028 study report version 1.0, 166 numbers traced to source, and a warning that 6 sections are not yet marked as reviewed — signing approves the whole report. She chooses the meaning “Study director approval — release to Quality Assurance for audit” and enters her password.
“The signature is linked to this exact version, with your name, the date and time and its meaning, and is kept in the audit trail.”
The report is locked as final v1.0 with all 9 sections approved, 166 numbers traced, 0 blanks and 0 gaps. It goes to Dr. Maya Chen for the Quality Assurance audit, and the export carries the number-to-source trace and the gap list, with how each was closed, as appendices.
“Signed — VLT-TOX-028 v1.0 sent to Quality Assurance for audit.”
The same study report, seen by the five people who carry it — what their fortnight looked like, and what it looks like now.
Six agents read, build, draft and check every report; a value the raw data lacks stays blank; the study director e-signs.
Reads the signed study plan and every signed amendment: the study design table, dose levels, group sizes for main, recovery and toxicokinetic animals, endpoints with time points, key dates, the guidelines followed and the sections the template needs.
Loads every table of the data package, maps it to its domain — BW, FC, CL, LB, OM, MA, MI, PC — and checks animal counts, days and groups against the study plan, recording any table the plan requires that is missing.
Builds every report table cell by cell from the source tables, with significance markers, keeping the house style's precision rules — nothing is typed by hand.
Writes each section in house style, describing test item-related changes by group and sex with direction, size and statistical significance. Every number goes in as a reference to its source cell, never free-typed.
Compares what the study plan requires with what the data package holds and what the template needs, and lists every missing file, empty cell, unsigned contributing report and exclusion that needs a statement.
Re-reads every number in the draft against its source cell, and checks that the summary and conclusion repeat the results exactly, that group sizes match the plan and that percentage changes are calculated from the cited cells.
Reviews, edits and e-signs. The agents propose; a named person decides.
Ask about a finding, a blank or what is holding up a signature — or say what to change. Every answer points to its source cell.
Summarise the 100 mg/kg/week findings in VLT-TOX-028
Male body weight gain 142.7 g against 156.8 g in controls (−9.0%); ALT 53 U/L against 38 U/L (+39%); liver weight 14.21 g against 12.84 g; minimal to mild centrilobular hepatocellular hypertrophy in 4 of 10 males; globulin up to 3.4 g/dL — the administered antibody. None looks adverse, but reversibility is open until the recovery histopathology arrives.
Why is a value blank in section 4.2?
Week 4 food consumption for Group 4 males is empty in the raw export — cell E5 of FC_Males. The study log shows Ana Kowalski weighed the food on 12 Sep, but that day's balance file was not transferred. The section drafter does not estimate, carry forward or average around a missing value, so it left a red blank and stated no 4-week mean.
What is blocking signature on VLT-TOX-028?
Two things: one value left blank — Week 4 food consumption for Group 4 males (FC_Males E5) — and missing recovery-group histopathology, which study plan section 9.4 requires for Groups 1R and 4R, so sections 4.5, 1 and 5 cannot state reversibility. There is also a minor item: a spilled Week 3 food weighing to confirm.
Add a rule: highlight any group mean more than 15% from control
Done. Every traced number in a draft that differs from its control by more than 15% is marked — in VLT-TOX-028 that is ALT (+39%) in 4.3, globulin (+19%, +31%) in the Summary and relative liver weight (+15%) in 4.4. You can change the threshold in Settings.
13 screens from the working solution, on its sample data. Pick one to see it large.
The template, house style, reviewers and the rules checked before every signature are settings. Two rules are required and cannot be switched off.
| Setting | Default | Choose from |
|---|---|---|
| Report template | Repeat-dose toxicity report v5.2 | Repeat-dose toxicity report v5.2 · Repeat-dose toxicity report v5.1 · Genetic toxicology report v3.1 · Safety pharmacology report v2.4 |
| Number style | Mean to 1 decimal, SD in brackets | Mean to 1 decimal, SD in brackets · Mean ± SD · Mean to source precision |
| Significance markers | * p<0.05 ** p<0.01 | * p<0.05 ** p<0.01 · a p<0.05 b p<0.01 · None |
| Pathology terms | Standard nomenclature (INHAND) | Standard nomenclature (INHAND) · As recorded by the pathologist |
| Signs the report | Study director of each study | Study director of each study · Dr. Lena Ortiz · Sam Patel · Dana Okafor |
| Missing raw values go to | Study technician of the study | Study technician of the study · Ana Kowalski · Report coordinator |
| Leave a value blank when it is not in the raw data | On — required | Never estimate, carry forward or average around an empty cell — cannot be switched off |
| Highlight differences from control | Off, threshold 15% | On or off · 10% · 15% · 20% · 25% |
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 studies; its by-hand figures are the team's reports before agents, as shown on the solution's dashboard · “target” = the design goal, measured in the live solution · “estimated” = our estimate. People, products and studies named on this page are fictional — characters and sample data in the working solution.
Software that drafts a GLP study report from the study's own records instead of having scientists re-key them. Preclinical Study Report Automation reads the signed study plan, the raw data export and the report template, drafts every section with each number linked to the cell it came from, and lists every blank and gap for the study director, who reviews, edits and e-signs.
Every number in the draft is inserted as a reference to its file, sheet and cell, never typed. Pointing at a number lights its cell beside the report, a calculated percent change lights both cells, and the consistency checker re-reads every number against its source before signature.
The agents leave a red blank — they never estimate, carry it forward from the week before or average around it — and the blank blocks signature. It is closed by a corrected, versioned source file from the person who holds the record, or by the study director stating it in the report, for example “not recorded”.
No. The section drafter describes test item-related changes by group and sex with their size and significance, and proposes adversity and the NOAEL for the study director to decide. Histopathology is reviewed by the study pathologist before the study director.
In the working solution: repeat-dose, genetic toxicology, safety pharmacology, reproductive toxicology, toxicokinetic and acute studies, each written to its own template — for example the repeat-dose toxicity report v5.2 or the genetic toxicology report v3.1.
A new or corrected file redrafts only the sections it feeds, the new numbers are re-checked against their cells, and the study director re-reviews the changes. In the sample study, the recovery histopathology redrafted sections 4.5, 1 and 5.
Yes. Raw data is read-only, agents draft and check but never sign, and an agent never fills a blank with a guess. The study director approves each section, edits as tracked changes and signs with a password and a stated meaning, linked to the exact version, before the draft goes to Quality Assurance.
The Agentic Solution Engine builds and deploys it from your requirements and documents — your report templates, house style and a sample of past study plans, data packages and signed reports — and it goes live once every quality gate has passed. We will walk you through it on your own studies first.
Sourced answers from papers, patents and reports in seconds, our own disagreements first.
1-min demoCompetitive Pipeline TrackingEvery rival trial, readout and approval in a sourced brief each Monday
1-min demoPost-Merger R&D IntegrationA signed, cited dossier for each asset before the experts leave.
1-min demoWe'll run Preclinical Study Report Automation on a sample of your own past data packages.