Ingest
Raw instrument files are frozen on receipt. Every input is hashed and listed in the manifest.
sha256 · 38 inputs
The pipeline
Five stages, one versioned run
Every dataset goes through the same stages with the same detector, the same measurement set and the same QC criteria. The version and every parameter are written to the manifest.
Detect
Objects are segmented from the height field with a fixed, versioned detector. No hand-picking.
1 903 candidates
Measure
Every object gets the same measurement set: height, volume, length, position, morphology.
per-object CSV
QC
Detections are reviewed against fixed criteria. Exclusions are counted, never silently dropped.
61 excluded · reasons logged
Report
Distribution, statistics and a methods section. Versioned, and traceable to each input file.
PDF · CSV · JSON
Raw height image Detected + measured
$ labtools analyze --pipeline particle-size@2.4.1 ./B2_run/ ✓ ingest 38 files · 38 hashes recorded ✓ detect 1 903 candidates ✓ measure height · volume · diameter ✓ qc 61 excluded (edge 44 · fused 17) ✓ report report_v1.pdf · objects.csv { "pipeline": "particle-size@2.4.1", "inputs": 38, "objects": 1842, "excluded": 61, "params": { "min_px": 4, "flatten": "plane" }, "stats": { "mean_nm": 31.4, "sd_nm": 9.7 }, "inputs_sha256": ["9f1c…", "a3d0…", …] }
Drag the divider to compare the raw height field with the analysed overlay. Values are schematic.
Analyses
What the pipeline measures today
Three sample types have a validated pipeline with fixed measurement rules, exclusion criteria and QC galleries. Anything else is assessed individually.
Validated pipeline
Nucleosomes or chromatin on DNA
Measures: Wrapped length, free arm lengths, nucleosome height and volume, nucleosome versus chromatosome, positioning on the DNA.
You get: Per-nucleosome table, wrapping and arm-length distributions, QC galleries of every traced molecule.
Tell us: The DNA construct (length in bp, positioning sequence) and the histones or linker histone used.
Validated pipeline
A protein bound to DNA
Measures: Fraction of DNA bound, copies per molecule, bound protein volume and height, binding position along the DNA, DNA bend angle.
You get: Per-molecule and per-grain tables, bound fraction with its control-calibrated false-positive rate, position histogram, QC galleries.
Tell us: The construct and binding-site positions, the protein and its expected size. A DNA-only image is the most useful extra file you can send.
Validated pipeline
Amyloid-beta aggregates
Measures: Aggregate height and volume, size distribution, morphology and elongation, surface loading, oligomer size in monomer equivalents.
You get: Per-aggregate table, height and volume distributions, field-level loading, particle crop galleries and traced fields.
Tell us: Peptide concentration, substrate (bilayer composition or mica) and incubation time.
Assessed individually
Something else
Measures: Particles, vesicles, fibrils, other DNA or protein samples. These are looked at individually before anything runs.
You get: Agreed case by case.
Tell us: What the sample is and what you need measured, as precisely as you can.
Quotes start from your data
Send one to ten images with your quote request
A quote request includes a few representative raw images, so the reply is about your data rather than data in general. They are checked by a person, run through the pipeline for your sample type, and the QC galleries are reviewed by hand before you get a reply.
- Raw instrument files: Bruker NanoScope, Asylum IBW, NT-MDT, Gwyddion, JPK, Park, Nanosurf.
- Up to 10 images, 250 MB each, uploaded over HTTPS.
- Nothing runs until a person has looked at the request.
- Sending a request commits you to nothing; paid work is quoted before it starts.
The full dataset comes later, once a feasibility assessment is agreed; large transfers are arranged individually.
Why the numbers hold
Three things this buys you
Statistical power
Measurement across thousands of objects rather than the dozens a person can hand-measure. Distribution shape, heterogeneity and rare subpopulations only become visible at that scale.
Reproducibility
The same inputs and the same pipeline version produce the same numbers every time, with the parameters recorded alongside the result.
Defensibility
Every reported number traces back to the image it came from. A reviewer, a collaborator or an auditor can follow the path from raw file to reported figure.
Process
How it works
Engagements run in two stages. The first exists so that the second can be quoted at a firm price, on work that otherwise varies considerably from job to job.
-
Stage 1: Feasibility assessment
You send 10 to 20 representative images. The pipeline is run against them to establish whether the measurement can be made on this sample type. You receive a short written assessment covering whether the measurement is possible, the expected number of measurable objects, any problems with the data, and a firm fixed-price quote for the full run.
You find out cheaply whether your data can support the question you are asking, before committing to a full project.
-
Stage 2: Full analysis
Begins once the quote is accepted. Inputs are frozen on receipt. The pipeline runs with its version and parameters recorded, detections are reviewed against fixed QC criteria, and the deliverable package is produced. Results reference the input manifest, so any number in the report traces back to a specific file.
Method development and training engagements are scoped individually and do not follow this two-stage path.
Scope
What a project includes
Work scales with the number of objects measured rather than the number of files, so one project is defined by a ceiling on both, whichever is reached first.
One project covers up to 40 images or the object ceiling stated in the quote, whichever is reached first. Datasets beyond this are quoted as multiple units, or at a per-image rate set in the quote.
Included in a project
- One measurement set as defined in the accepted quote.
- QC review against fixed criteria.
- The distribution and summary statistics.
- The PDF report.
- The per-object CSV.
- The run manifest.
- One round of revisions covering clarification and presentation.
Quoted separately
These are not included. They are listed here so the boundary is clear before a project starts rather than after.
- Additional measurands not in the original quote.
- Re-running after additional images are supplied.
- Statistical comparison against conditions not named at quoting.
- Figure restyling for journal submission.
- Response-to-reviewer analysis.
Data that cannot be measured
Some images cannot be measured. The usual reasons are:
- Resolution falls below the pixel-size floor.
- Object density is too high to separate reliably.
- Substrate contamination exceeds threshold.
- Files are damaged or incomplete.
Where a defined proportion of a dataset fails, delivery proceeds on the measurable portion and the remainder is reported rather than discarded. The exclusion count and the criterion that produced it appear in every report.
Exclusions are always counted and always reported. Data is never silently dropped or filtered. Where more than 10% of a dataset cannot be measured, work pauses and you are consulted before it continues, rather than a partial result being delivered as though it were complete.
Turnaround
Projects are delivered within one week of receiving the complete dataset. Turnaround is quoted from receipt of the complete dataset, not from first contact, and the clock pauses while awaiting decisions or clarification.
Revisions
One round of revisions covering clarification and presentation is included. Reports are versioned, so a revised report is issued as a new version and never overwrites the one you already have.
What you receive
Deliverables
Every project returns three artifacts.
versioned
PDF report
Figures, summary statistics, and the methods section describing what was measured and how.
every measurement
Per-object CSV
Every individual measurement, not just the summary. You can reanalyse the distribution yourself, apply your own cutoffs, or combine it with other data.
provenance
JSON run manifest
Input file hashes, pipeline version, every parameter used, and every exclusion with its reason. This is what makes a result checkable months later.
Reports are versioned. A revised report is issued as a new version and never overwrites the one you already have.
Inputs
Formats and requirements
Accepted formats
Tested end to end:
Bruker NanoScope Asylum Research IBW NT-MDT Gwyddion (.gwy)
Most other instrument formats can be supported. If yours is not listed, say so in your enquiry. Whether it can be read is confirmed during the feasibility assessment, before you commit to a full project.
Image requirements
Measurability is set by object size relative to pixel size rather than by any absolute figure. As a working rule, an object needs to span roughly four to five pixels across before its size can be measured reliably, so pixel size should be no larger than about a quarter of the smallest feature of interest. Below that, the measurement floors out and small objects are lost rather than mismeasured.
Raw or minimally processed data is preferred, with any flattening already applied recorded. Height measurements depend on the background definition, so an undocumented flattening step cannot be undone or accounted for.
What cannot be measured
What can and cannot be measured depends on the measurement being requested and on the quality of both the imaging and the sample. The feasibility assessment exists to answer that question against your actual data rather than in the abstract. Some cases are known in advance:
- Exported PNG or TIFF images. These carry a picture of the surface rather than the height data every measurement here depends on. Raw instrument files are needed.
- Objects below the pixel-size floor for the requested measurement. They are not mismeasured, they are not detected at all, and their absence shifts the distribution.
- Fields crowded enough that touching objects cannot be separated reliably. Fused neighbours inflate both the large end of a size distribution and the apparent elongation of the objects in it.
- Absolute height where no flattening record exists. Relative comparisons within a consistently processed dataset remain valid.
- Comparisons pooled across substrates or imaging sessions with different backgrounds and detection floors. These are reported separately rather than combined.
Ways to work together
Engagement types
Standard path
Project analysis
One condition set measured against an agreed definition, returned as a report with the per-object data and run manifest. The two-stage path above applies.
Scoped individually
Method development
Building and validating a measurement pipeline for a specific assay where none currently exists. Scoped individually, with validation criteria agreed before work starts.
By format and location
Training
Workshops on quantitative AFM analysis, covering measurement definition, detection review, and how to report distributions defensibly. Quoted by format and location.
Pricing
A structure and a floor, not a rate card
Jobs differ in object count, image quality, sample type and the number of measurands requested, so a published table would be wrong for most visitors. The feasibility assessment exists so that a firm fixed price can be given once the data has been seen.
| Engagement | Price | Notes |
|---|---|---|
| Feasibility assessment | $500 | Credited in full toward the project if you proceed. |
| Project analysis, academic | from $1,500 | Firm fixed price after assessment. |
| Method development | from $4,000 | Validation criteria agreed before work starts. |
| Training and workshops | $1,200 per day | Quoted by format and location. |
| LabTools, academic and individual | Free | In active development. |
| Core facility site license | On enquiry | Institutional pricing. |
| Commercial and industry work | Quoted | After assessment. |
All projects are subject to a minimum of $1,500. Published figures are starting points subject to assessment, not fixed rates.
Full pricing sheet, with a printable PDF
What changes a quote
Three broad categories, in increasing order of effort:
- Standard
- The measurement is one the pipeline already performs on this sample type.
- Non-standard
- The pipeline requires parameter tuning or adaptation for the sample.
- Development
- The measurement does not yet exist and must be built and validated. Scoped as a separate engagement.
What the estimate is based on
- The number of objects to be measured.
- Whether the sample type is already supported.
- How much manual review the detections require.
- How many distinct measurands are requested.
Data handling
Confidentiality
Access-scoped storage
Client data is held in access-scoped storage. It is not used for any purpose other than the engagement without written agreement.
90-day retention
Raw inputs are retained for 90 days after delivery and then removed.
NDA on request
An NDA can be put in place on request before any data is sent.
Next step
Start a feasibility assessment
Tell me what you are trying to measure and send a few representative images. You will get a reply saying whether the measurement is feasible, what the feasibility assessment would involve, and what it would cost.