Skip to content

CHANGELOG.md

Latest commit

 

History

History
165 lines (151 loc) · 9.62 KB

CHANGELOG.md

File metadata and controls

165 lines (151 loc) · 9.62 KB
Raw

Changelog

Version history for the SPARCAL Spatial-SNV Viewer. The running version is sparcal_viewer.__version__; the About dialog (click the app name on the top bar) shows it alongside the build time.

1.6.0

  • Eight paired P4/P6 CalicoST-ablation studies. Each replicate now has a distinctly named full_calicost somatic profile and a spatial_only_no_calicost profile, so both arms can be loaded side by side without file or study-name collisions.
  • The tracked study manifest records hg19 matrix dimensions, actual voting features, whether CalicoST purity was informative, archive sizes, and SHA-256 checksums. P4 replicate 2 is explicitly marked as zeta,theta because its purity estimates are uninformative.

1.5.0

  • Auto tumor regions: new "Method" selector. raw_burden detects tumor extent from un-normalized SNV burden — validated on the DCIS sec2 ground truth as the best extent detector (raw burden ≈ per-spot coverage/cellularity, which tracks the dense DCIS ducts; coverage-normalizing it removes that signal). customized keeps the prior fully tunable behaviour, including the optional "Normalize by coverage (UMI)" checkbox. Default is raw_burden; selecting it forces normalization off and disables the checkbox, while customized re-enables it (checked when a study has coverage). A profile / latent-factor method for clonal substructure is planned separately.

1.4.0

  • Colour-blind-safe Overview palette. The Overview / profile-map export no longer cycles the full HSV rainbow (pg.intColor), which put a pink region hue right next to the grey "multiple regions" marker — indistinguishable under colour-vision deficiency. Regions now use a fixed 8-hue categorical palette validated CVD-safe (worst adjacent ΔE 24.2), and the multiply-assigned marker is now near-black (#1a1a19) so it can never collide with a region hue. With >8 regions the palette repeats; position + the region list keep them distinguishable (colour is no longer the sole cue).

1.3.0

  • New "Overview" button (left view, directly under Reset): paints every tumor region in the current profile its own colour so the whole profile is visible at a glance. Spots claimed by more than one region are shown dark grey; spots in no region stay pale. Hover still names the region under the cursor, and Reset clears the overview.
  • Fix: turning off "color by SNV count" now restores a uniform colour. Spots kept their per-spot burden colours because pyqtgraph's single-brush setBrush leaves stale per-point brushes in place; the view now clears them, so toggling the mode off (and Reset / Overview) repaints every spot uniformly.

1.2.0

  • Magnitude-aware burden — gene-expression studies now carry real expression. The per-spot burden is now the row-sum of the matrix values instead of a count of nonzero columns. For the binary SNV studies this is identical (values are 0/1), but a magnitude matrix (e.g. a gene-expression study) now drives the burden/Auto-region signal by expression level, not mere gene presence.
    • DLPFC_151507_GEX.zip was rebuilt: values now encode log-normalized expression — normalize_total(1e4) per spot → log1p → round(x * 36) → uint8 (0 = absent, preserved exactly).
    • The "% of spots" variant grouping (generate_*) and the cached per-feature total now use a presence count (value > 0), so they stay correct regardless of whether the matrix stores presence or magnitudes.

1.1.0

  • New example studies on the SPARCAL-studies-v1 release:
    • DCIS_1_SPARCAL.zip — DCIS section 1, SPARCAL SNV calls (merged germline + UPV + somatic matrix).
    • DCIS_2_SpatialSNV.zip — DCIS section 2, SNVs called by SpatialSNV (Mutect2-based); same Visium section as the bundled DCIS_2_SPARCAL.
    • DLPFC_151507_GEX.zip — DLPFC 151507 gene-expression study. The matrix is a per-spot gene-presence matrix (gene detected → 1) from the Visium filtered_feature_bc_matrix, so per-spot "burden" is gene complexity and the SNV list / Auto regions operate on genes.
  • P4/P6 now ship coverage: P4_rep1_SPARCAL.zip and P6_rep1_SPARCAL.zip now include a spot_coverage.csv, so Auto tumor detection coverage-normalizes the SNV burden for those sections instead of falling back to raw burden.
  • Every new/updated study bundles a spot_coverage.csv (per-spot total UMI).

1.0.1

  • Auto tumor regions — every knob is now a slider + spin box: Intensity, Grow margin, Min region size, and Split valley depth each pair a drag slider with a spin box, so you can drag, step with the up/down arrows, or type a number; the two stay in sync. New defaults: Grow margin 10 (was 30) and Split valley depth 40% (was 30%).
  • About dialog now reachable: the app name on the top bar was a bare menu-bar action, which does not render on macOS's native menu bar. It's now a proper app-name menu whose "About …" item (folded into the application menu on macOS) opens the version/build-time dialog.

1.0.0

  • Hover hit-radius fix: the region label now appears even when the cursor is in the gap between spots — the snap radius is derived from the spot pitch instead of a fraction of the spot size.
  • Multiple centers per Auto region: an Auto region keeps all its seed centers (not just the strongest), so a basin fused from several peaks shows several stars — exposing how many potential sub-regions a collection contains.
  • Export profile map: File ▸ Export ▸ "Profile map (with/without background)" renders every region of the current profile in its own colour to PDF (vector) or PNG (by file extension).
  • Compare profiles: on the profile page, Edit → select two profiles → Function ▸ "Compare profiles" shows a region×region Jaccard overlap table plus cluster-agreement scores (ARI, NMI, homogeneity, completeness, V-measure), all computed pure-numpy. Item set = all in-tissue spots, unassigned → background.

0.5.0

  • Tumor profiles: Regions are now organized into named profiles (separate separations of the tissue — e.g. "Ground Truth", "Test", auto-detected). Column 2 has a new top layer: pick a profile, then its regions; the region view is titled Tumor profile: "<name>" with a ‹ Profiles back-button. New/Rename/Delete profile management. Persistence adds a profile column to tumor_groups.csv, tumor_centers.csv, and variant_groups.csv; files without it load into a single Default profile (back-compatible). DCIS_2 ships as Ground Truth (T1–T11), Test (Test_1–Test_16), and Other (ALL, T3Eexclusive).
  • Reset: new top-left button on the spatial view clears all region/group/SNV selection and paints every spot a uniform pale white.
  • Auto regions — clearer view: opening the Auto dialog resets the tissue so the coloured region preview reads against a blank background, and the dialog now always floats on top of the main window.
  • Hover-to-identify: in normal mode, hovering a spot shows its region name as a label at the cursor and selects/shows that region in column 2 (live, updating only when the hovered region changes).

0.4.0

  • Region/group edit selection: Edit mode now uses extended selection so Shift-click selects a contiguous range and Ctrl/Cmd-click toggles individual regions or groups (previously single-toggle only, with no working Shift range-select).
  • About dialog: clicking the app name on the top bar opens an About box showing the version and build time. Build time is stamped into _build_info.py by build_macos.sh, falling back to the source mtime when running from a checkout.

0.3.0

Watershed regions, selection legend, saved centers.

  • Auto tumor regions now use a seeded watershed: two centers that grow into each other fuse only across a shallow saddle (new "Split valley depth" control) instead of always merging on contact.
  • Showing a group / selected SNVs on tissue colours spots by how many of the selected SNVs each covers, with a selection-specific legend (top-left).
  • Auto regions save their center (strongest seed) to tumor_centers.csv and mark it with a star when the region is selected.
  • Packaging: build_macos.sh bundles the DCIS_2 study inside the .app and the app opens that bundled config by default (copied to a writable per-user dir when frozen so edits persist); tools/publish_release.sh uploads built zips to a GitHub Release.

0.2.0

SNV-burden spot coloring and auto tumor-region detection.

  • Colour on-tissue spots by per-spot SNV count on a sky-blue→purple quantile ramp, with a count legend ("color by SNV count" toggle).
  • "Auto" tumor regions via hysteresis-thresholded seeded region growing on the Visium grid: intensity slider, grow margin, min region size, optional coverage normalization, and manual seed add/exclude by lasso.
  • Optional per-spot spot_coverage.csv (barcode,total_umi) for coverage normalization; tools/make_spot_coverage.py generates it from raw SpaceRanger output.

0.1.0

SNV variant file converted to JSON; file management upgrade.

  • Export SNVs as .json with contents/source/variants structure tracking the originating region and group for provenance.
  • Import reads .json and navigates to the matching region/group, re-creating them from the variant list if they no longer exist.
  • Edit mode supports deleting individual variant groups (not just whole regions), with a confirmation dialog listing everything affected.
  • Column-3 action bar always visible (Show on tissue / Add spots); Import button added; empty spatial pane shows a "Click to open a config file" prompt.