Fewer runs.
Ranges you can defend.
Every development run costs money and, more expensively, calendar time. Every decision you take has to be defended twice, once at the stage gate and again in the filing, years later, by someone who wasn't there. The platform keeps the design, the data and the conclusion attached to each other so neither of those costs compounds.
The unit of work
A study, with its design, runs, results and conclusion in one object
The analysis
Regression, correlation, PCA and DoE on the run data itself. No export
The output
A development report drafted from the studies, with every figure bound to source
From the workflow down to the run data.
The process development workflow is the map. Open a stage, open a unit operation, open a study, and you land on the runs themselves, with the charts and the analysis already built.
process developmentProcess development workflow
Edit workflowUpstream process development (Mab) workflow
Create upstream development reportStudies
Study 1: Temp and pH shift
10 runs · complete
Study 2: Feed study
8 runs · in progress
Equipment
- 10 L Sartorius bioreactorBRX-0010-03 · in calibration
- Cedex cell counterCDX-4402 · in calibration
- Mettler Toledo pH meterPHM-1188 · due in 21 days
- Nova Biomedical bioanalyzerNVA-7731 · in calibration
Materials
- Basal mediumBM-2401-114
- Feed AFEED-M-2402
- Base, 1 M Na₂CO₃BS-2312-009
- Antifoam CAF-2401-052
Process parameters
| Parameter | Target range | Value |
|---|---|---|
| Working volume | 6 – 10 L | 8 L |
| Seeding density | 0.3 – 0.5 ×10⁶/mL | 0.4 ×10⁶/mL |
| Starting pH | 6.8 – 7.4 | 7.3 |
| pH deadband | ± 0.03 – 0.08 | ± 0.05 |
| Temperature shift | 32 – 34 °C | 34 °C |
| Temperature shift day | 6 – 8 days | 7 days |
| Dissolved oxygen | 20 – 50 % | 40 % |
| Agitation | 80 – 150 rpm | 100 rpm |
| Air sparging | 0.01 – 0.1 vvm | 0.02 vvm |
| Feed start | day 3 – 4 | day 3 |
| Feed rate | 3 – 5 % v/v daily | 3 % daily |
| Osmolality | 280 – 320 mOsm/kg | 298 mOsm/kg |
| Antifoam | 0.01 – 0.05 % | 0.02 % |
| Harvest criterion | viability < 80 % or day 14 | day 14 |
In-process controls
- Viable cell density, viabilitydaily
- Glucose, lactate, glutamine, ammoniadaily
- Osmolalityevery other day
- Offline pH, pCO₂daily
- Titerday 7, 10, 12, harvest
Procedure
written once · executed ten times · two values assigned per run
- 1Seed at 0.4 ×10⁶ cells/mL, 36.5 °C, pH 7.10
- 2Feed A at 3 % of working volume daily from day 3
- 3Shift temperature to 33 °C on the assigned day
- 4Hold pH at the assigned setpoint from that day
- 5Harvest at viability below 80 %, or day 14
Run data
recorded against the run · nothing re-keyed
| Run | Shift day | pH | Peak VCD | Titer g/L | Viability |
|---|---|---|---|---|---|
| Run 1 | 3 | 6.90 | 19.0 | 3.36 | 82 % |
| Run 2 | 3 | 7.10 | 19.7 | 3.45 | 81 % |
| Run 3 | 4 | 7.00 | 20.4 | 3.74 | 81 % |
| Run 4 | 5 | 6.95 | 21.1 | 3.84 | 80 % |
| Run 5 | 6 | 7.00 | 21.8 | 3.94 | 80 % |
| Run 6 | 6 | 7.20 | 22.5 | 3.59 | 79 % |
| Run 7 | 7 | 6.90 | 23.2 | 3.76 | 78 % |
| Run 8 | 8 | 7.05 | 23.9 | 3.80 | 78 % |
| Run 9 | 9 | 7.00 | 24.6 | 3.62 | 77 % |
| Run 10 | 9 | 7.15 | 25.3 | 3.43 | 77 % |
Harvest titer over culture duration
titer g/L · culture day on the x axis
Online pH over culture duration
online pH · the shift day is visible per run
Viable cell density and viability
median with the spread across all ten runs · viability dashed
Titer across the two factors the study varied
shift day on the x axis · pH setpoint on the y axis
Ten runs, plotted from the data the operators recorded. Nothing was re-keyed to get here, and the same numbers feed the study report.
Principal component analysis shows how the runs cluster. Draw a region on the plot and the platform turns it into recommended ranges.
What the data says
- Shift day and pH setpoint together explain most of the spread in titer10 runs · response surface
- The best predicted condition is a day-6 shift with pH held at 7.00Runs 5, 8 · 3.94 and 3.84 g/L
- Past day 8 the gain is given back through viability at harvestRuns 9, 10 · viability 77 and 76 %
Recommended next steps
- 1Three confirmation runs at day 6, pH 7.00The optimum is interpolated, not executed. Nothing moves forward on a predicted point.
- 2Narrow the pH range to 6.95 – 7.05 for study 2Outside that band the model loses more than 0.2 g/L.
- 3Carry feed rate in as the next factorIt was held at 3 % throughout, so its effect is unmeasured here.
The cost you are already paying.
Design lives in one place, data in another, the conclusion in a third
The DoE is in a spreadsheet, the runs are in a historian or a runsheet, the interpretation is in a slide deck, and the report is written months later from all three. Every handoff loses context, and the reconciliation work at the end is where the schedule actually goes.
The analysis queue is a bottleneck
A scientist with a question about their own campaign waits on a statistician or a data team. The question is usually simple, is this correlated, does this factor matter, is this run an outlier, but the round trip is measured in days, so most questions go unasked.
Rationale evaporates
Six months later nobody can say why a range is what it is. The stage gate asks, and a senior scientist spends a week reconstructing an argument that was obvious at the time.
The report is the last mile, and it is the longest
Development reports are assembled by your most expensive people re-typing numbers out of spreadsheets, then reconciling the version that got circulated against the version that got signed.
On the platform.
Design and result in one object
A study carries its design, its runs, its result tables and its conclusion. Change the data and everything downstream: the analysis, the charts, the report section, recomputes rather than drifting.
Ask your own data directly
Plain-language questions against your result tables. The platform picks the analysis, runs it, and you get a chart and an interpretation in the meeting rather than the one after it.
Statistics that are part of the record
Regression, correlation, PCA, outlier detection and DoE run against the same data the report will cite, so the number in the chart and the number in the document cannot disagree.
Rationale captured where the decision was made
The reason a range was set sits with the study that set it, linked to the runs that justified it. When the stage gate asks, the answer is a query rather than an archaeology project.
Reports drafted from the record
Your approved template governs the structure; the platform binds the figures. The author edits prose instead of transcribing numbers, and a regeneration picks up any data that has changed.
Multi-modality without a fork
The model is built on unit operations, parameters, criteria and results, so a mAb study, an AAV study and an mRNA study sit on the same system with the same tooling.
Related capabilities
Bring us the work you already know the answer to.
We would rather you judge the platform against a result you can already check than sit through a scripted tour.
Book a demo