This document maps the user testimonies gathered during discovery to concrete product requirements for Parallax.
The goal is simple: separate what is truly required for a viable FlowJo alternative from what is merely nice to have.
The testimony set is remarkably consistent. Researchers rely on FlowJo for:
- daily gating and review of flow cytometry data
- compensation review and correction
- dot-density, histogram, and other plot views with adjustable scaling
- batch application of gating strategies and templates across many samples
- quantitative statistics, formulas, and publication-ready exports
- saved, reproducible analysis workflows that can be reopened later
- working away from acquisition instruments so cytometers are freed for other users
The strongest use cases in the testimony set include:
- immunophenotyping and immune-cell subset analysis
- cell cycle, proliferation, apoptosis, and DNA-damage assays
- intracellular staining and inflammatory marker analysis
- CRISPR editing, screening, and engineered-cell workflows
- mitochondrial function, viral infection rate, and membrane-protein quantification
- longitudinal and clinical-trial style batch analysis
Parallax already has a real analytical foundation:
- deterministic Rust command log and replay
- rectangle and polygon gates
- population hierarchy with parent-child gating
- undo and redo on explicit command state
- FCS parsing in Rust and CLI inspection
- authentic real-world parser regression suite
That foundation is valuable, but it is not yet enough for real lab adoption.
| Capability | Evidence From Testimonies | Why It Is Must-Have | Current Status In Parallax | Priority | Recommended Epic |
|---|---|---|---|---|---|
| Desktop FCS import and sample browser | SePil Lee, Chad Williamson, Adriana Golding, Sitanshu Sarangi all describe opening acquired files and analyzing them away from the instrument | Without direct file import, the desktop cannot replace FlowJo in daily work | Partial. Desktop users can now import selected .fcs files, recursively import .fcs experiment folders in deterministic order, and switch samples, but richer import review, saved import manifests, and downstream batch workflows are still missing |
P0 | EPIC A |
| Multi-sample workspace and session management | Jack Yanovski, Deliang Zhang, Chad Williamson, Sitanshu Sarangi describe workflows across many experiments and time points | Real labs do not analyze one sample at a time in isolation | Partial. One local session can now hold multiple imported samples with per-sample command history, persisted workspaces, early batch template application, grouped stats export, selected-population comparison, persisted cohort labels, and structured sample metadata, but sample-sheet merge and richer cohort organization workflows are still missing | P0 | EPIC A |
| Compensation review, edit, and application | SePil Lee explicitly calls out detecting improper compensation and correcting it immediately; Jack Yanovski names compensation as a core function | Compensation is foundational to trustworthy flow analysis | Partial. The desktop can now inspect compensation QC, apply the parsed FCS matrix, paste a manual spillover override, clear the override, and replay those choices deterministically, but matrix-grid editing and reference validation against external controls are still missing | P0 | EPIC B |
| Plot system beyond scatter plots | SePil Lee, Keita Saeki, Deliang Zhang, Marcela Teatin Latancia all depend on richer visual analysis | A Flow workstation needs density, histogram, and flexible visualization, not just scatter | Partial. Scatter, histogram, and density-grid views exist, but user-authored plot layouts, contours, and publication styling are still missing | P0 | EPIC C |
| Axis scaling and transforms | SePil Lee mentions customizable axis scales; many assays require logicle or biexponential transforms for interpretation | Marker intensity analysis depends on the right transform model | Partial. The desktop now supports replayed Linear, Signed Log10, Asinh, Biexponential, and Logicle transform presets, plus explicit Auto, Focus, zoom, pan, and manual range plot-view controls, but it still lacks fully tunable reference-matched transform implementations |
P0 | EPIC B |
| Saved workspaces | Marcela Teatin Latancia, Jack Yanovski, Deliang Zhang, Chad Williamson all emphasize reproducibility and reuse | Reopening analysis state is table stakes for real use | Partial. The desktop can now save and reopen source-path workspaces and portable .parallax bundles that copy FCS sources and validate sample fingerprints, compensation hashes, acquisition metadata hashes, and source-byte integrity, but derived caches, recovery snapshots, compression, and signed package manifests are still missing |
P0 | EPIC D |
| Gating templates and batch application | Marcela Teatin Latancia, Jack Yanovski, Deliang Zhang, Sitanshu Sarangi all describe repeated analysis across many files | This is one of the clearest speed and reproducibility advantages FlowJo has today | Partial. The desktop can now save/load named gate-template files, apply the active sample's gate template across other loaded compatible samples, compare the resulting population across samples and cohorts, and export grouped stats, but richer grouping, channel-alias mapping, and template-library management are still missing | P0 | EPIC D |
| Statistics engine | Marcela Teatin Latancia, Keita Saeki, Anup Dey, Tina Maio need quantitative outputs, not only plots | Researchers need counts, frequencies, central tendency, and assay-specific summaries | Partial. The desktop now computes counts, parent/all frequencies, per-channel mean/median/gMFI/P5/P95/CV summaries, positive-fraction metrics, median-signal metrics, gMFI-signal metrics, stain-index metrics, and mean-ratio metrics, but broader assay summaries are still missing | P1 | EPIC E |
| Custom formulas and derived metrics | SePil Lee explicitly mentions customized formulas; Tina Maio describes converting signal into quantitative metrics | Labs need derived values, not only raw gated counts | Partial. The desktop now supports replayable positive-fraction, median-signal, gMFI-signal, stain-index, and mean-ratio metrics on the selected population, shows them across samples and cohorts, and exports them as CSV, but there is still no free-form expression editor or spreadsheet-style formula layer | P1 | EPIC E |
| Export to CSV and Excel | SePil Lee and others need direct table export for downstream analysis and reporting | No export means no practical downstream workflow | Partial. The desktop can now export active-sample stats, selected-population comparisons, cohort summaries, derived metrics, and grouped batch stats as CSV with structured sample metadata columns, but Excel export and report-ready layouts are still missing | P1 | EPIC F |
| Publication-quality figure export | Jack Waite, Deliang Zhang, Anup Dey, Chad Williamson, Guillaume Gaud all mention figure generation for meetings and manuscripts | Publication-grade output is a central reason people use FlowJo | Partial. Plot cards can now export high-resolution PNG figures, provenance-footed page PDF figures, and a provenance-footed single-page plot report PDF with controls hidden, but SVG/vector export, richer multi-panel layouts, style presets, and report packaging are still missing | P1 | EPIC F |
| Comparison workflows across samples and conditions | Jack Waite, Jack Yanovski, Antony Cougnoux, Guillaume Gaud, Tina Maio all compare treated vs control, time points, or longitudinal cohorts | A single-sample tool is insufficient for modern biology workflows | Partial. The desktop can now compare the selected population across loaded samples, assign cohort labels and structured sample metadata, aggregate cohort-level summaries, and export those views, but it still lacks manifest merge, review-by-exception triage, and richer cohort visual layouts | P1 | EPIC G |
| Gate editing, quadrant gates, and navigation polish | Daily users need to refine gates precisely and quickly across projections | Initial gate creation alone is not enough for expert analysis | Partial. Rectangle, polygon, quadrant, and histogram range gates now create replayable populations, selected rectangle, polygon, and histogram range gates can be refined through append-only exact fields or draggable plot handles, and plot views support auto/focus/zoom/pan actions, but richer navigation affordances are still missing | P1 | EPIC H |
| Parser compatibility with real instrument output | Deliang Zhang cites broad file compatibility; Pedro Pereira Da Rocha points out lack of viable alternatives | If authentic cytometer files fail to load, trust collapses immediately | Partial. The authentic suite now passes 39/39 pinned public files under --require-all-pass, but the corpus still needs to grow toward 100+ files across more vendors and edge cases |
P0 | EPIC I |
| Reproducibility and audit trail at workspace level | Marcela Teatin Latancia, Jack Yanovski, Deliang Zhang, Chad Williamson all emphasize consistency and reproducibility | The command log is a strong start, but users need saved and exportable analysis lineage | Partial. Command replay exists, but workspace persistence and report export do not | P1 | EPIC D |
The testimonies point to a very clear order of operations.
- desktop FCS import
- multi-sample workspace model
- compensation workflow
- transforms and axis scaling
- histogram and density plots
- workspace save/load
- batch templates and template application
- parser compatibility expansion from the current
39/39authentic public-file gate toward a broader100+file corpus
- statistics engine
- custom formulas
- CSV and Excel export
- publication-quality figure export
- comparison workflows
- gate editing, quadrant gating, and navigation polish
- cloud sync
- background jobs
- AI copilot
- retrieval and semantic workspace search
Deliver:
- import one or many FCS files from disk
- sample list and sample switching
- session model for multi-sample analysis
Acceptance:
- open a folder of FCS files in the desktop
- switch between samples without restarting
- preserve selection, gating tree, and sample context in one session
Deliver:
- inspect parsed compensation matrices
- apply or override compensation
- linear, logicle, and biexponential transforms
- axis scaling controls
Acceptance:
- compensation changes update plots and populations deterministically
- transformed axes remain stable across sessions and replays
Deliver:
- scatter
- dot density
- histogram
- density view
Acceptance:
- users can switch plot type per panel
- axis settings and transforms are visible and adjustable
- interaction remains smooth on large event counts
Deliver:
- save/load workspace
- reusable named gating templates
- apply template across groups of samples
Acceptance:
- a saved workspace reopens identically
- template application across N samples is deterministic
- loaded templates validate before replacing a sample's gate history
- command-log lineage remains preserved
Deliver:
- count
- frequency
- mean, median, gMFI, percentiles, and CV
- positive fractions
- median signal
- gMFI signal
- stain index
- custom formulas over gated populations
Acceptance:
- results are reproducible and exportable
- formulas recalculate correctly after gating changes
Deliver:
- CSV and Excel table export
- figure export for plots and layouts
- PDF-ready outputs
Acceptance:
- exported numbers match on-screen analysis
- exported figures are publication-ready without manual recreation elsewhere
Deliver:
- compare samples side by side
- compare grouped conditions and time points
- summary tables across groups
Acceptance:
- users can analyze control vs treatment and longitudinal cohorts directly in Parallax
Deliver:
- editable gates
- quadrant and histogram range gates
- pan and zoom
- gate handles and refinement tools
Acceptance:
- users can iteratively refine a gate without recreating it from scratch
Deliver:
- close known parser gaps exposed by the authentic test suite
- expand authentic-file coverage as new sources are added
Acceptance:
- keep the current expected-failure count at zero
- maintain zero unexpected regressions in the authentic suite
If we want to move the product meaningfully toward the testimony-defined target, the next engineering sequence should be:
- EPIC A: desktop import plus multi-sample sessions
- EPIC B: compensation plus transforms
- EPIC C: histogram and density plots
- EPIC D: workspace save/load and batch templates
- EPIC I: parser hardening in parallel with the above
That sequence keeps the product focused on the real promise users care about:
- fast
- trustworthy
- reproducible
- practical for daily flow cytometry work
The testimonies do not justify spending early effort on:
- AI-first auto-gating
- fancy dashboards
- plugin marketplace work
- real-time multi-user editing
- browser-first UI at the expense of desktop performance
Those may matter later, but they are not the core job users are hiring the product to do.