8bb883ed32e24a1007bf9f97ee1793a46bf47bd0
mspeir
  Thu Aug 27 10:00:43 2026 -0700
singleCellSignalsPeaks: classify the SEA-AD and Allen tracks, add BrainVar stages

72 hg38 SEA-AD and 89 mm10 allen-basal-ganglia tracks had lost their broad cell
class and were drawing in dataset colours instead of the shared palette; both
assemblies are now 0 unclassified. Fix is in the hub build (cellBrowser
ucsc/allTracksHub).

Adds an "Other glia" class, since "Other" was doing four unrelated jobs. BrainVar
goes 4 -> 13 hg38 subtracks with the new stage-split pseudobulk. Scaling moves to
"autoScale group" on the composite so selected subtracks share one scale, and
every bigWig now carries a data range -- without one the 13 BrainVar tracks drew
as flat lines against hgTracks' built-in 0:127.

Description pages and makeDocs updated, including removing the stale claim that
SEA-AD is coloured by its own subclass palette.

refs #37914, refs #37820

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

diff --git src/hg/makeDb/doc/mm10/singleCellSignalsPeaks.txt src/hg/makeDb/doc/mm10/singleCellSignalsPeaks.txt
index 8e426b25a3c..69a7ad01e47 100644
--- src/hg/makeDb/doc/mm10/singleCellSignalsPeaks.txt
+++ src/hg/makeDb/doc/mm10/singleCellSignalsPeaks.txt
@@ -1,216 +1,302 @@
 # mm10 singleCellSignalsPeaks track  -  2026-07-22  Claude (mspeir)  refs #37914
 
 # The native mm10 "singleCellSignalsPeaks" faceted composite is the Genome
 # Browser version of the mm10 signal-&-peaks composite (cellBrowserMm10) of the
 # UCSC Cell Browser all-tracks super hub, whose build lives in the cellBrowser
 # repo, ucsc/allTracksHub:
 #   https://github.com/ucscGenomeBrowser/cellBrowser/tree/develop/ucsc/allTracksHub
 #
 # It gathers the per-cell-type ATAC-seq signal (bigWig) and peak (bigNarrowPeak) tracks from the
 # single-cell ATAC datasets in the Cell Browser and re-parents them under one
 # faceted composite. It is the mm10 counterpart of the hg38 track of the same
 # name (see doc/hg38/singleCellSignalsPeaks.txt, Redmine #37820). Histone marks,
 # cCREs, RNA/expression, interactions, and the separately-faceted motor-neuron
 # and brain-spatial sets each live in their own hub composites and are NOT part
 # of this track.
 
 ##############################################################################
 # 1. Source data
 ##############################################################################
 # The track mirrors the hub's main mm10 signal-&-peaks faceted composite
 # (cellBrowserMm10): 647 subtracks (641 bigWig, 6 bigNarrowPeak) from 9 datasets
 # - catlas-mouse-aging (234), catlas-mouse-brain (160), allen-basal-ganglia-atac
 # (131), olg-dyn-eae-multiome (28), mouse-lvcp-multiome (23, incl. the 6 peaks),
 # catlas-paired-tag (21), mouse-kidney-atac (19), olg-eae-ms (16) and
 # mouse-epi-juv-brain (15). That composite and its facet metadata are produced by
 # the hub build from the Cell Browser dataset tree:
 #
 #   cd $HOME/cellBrowser/ucsc/allTracksHub
 #   python3 build_manifest.py            # scan datasets -> manifest.tsv
 #   python3 build_stanzas.py             # manifest -> stanzas/mm10.trackDb.txt
 #                                        #            + meta/mm10.metadata.tsv
 #
 # The build writes its output to $CBHUB_OUT, NOT next to the scripts (that dir is
 # git-controlled). Default: /hive/data/inside/cells/all-tracks-hub-build
 #
 # The per-track source files (abs_path column of manifest.tsv) are the files the
 # Cell Browser datasets already serve; nothing is regenerated here, only copied.
 
 ##############################################################################
 # 2. Copy the data files into place  (bed dir, served via a /gbdb symlink)
 ##############################################################################
 # copySingleCellSignalsPeaksFiles.py copies every cellBrowserMm10 subtrack file
 # into /hive/data/genomes/mm10/bed/singleCellSignalsPeaks/<served-relpath>,
 # keeping each file's served relative path, and copies the composite's facet
 # metadata to <bed>/singleCellSignalsPeaks_metadata.tsv. The served subpath is
 # preserved on purpose: some coverage/peak basenames (e.g. MOL.bw, OPC.bw) repeat
 # across datasets, so a flat directory would clobber them.  89.6 GB, 647 files.
 #
 #   scriptDir=$HOME/kent/src/hg/makeDb/scripts/singleCellSignalsPeaks
 #   python3 $scriptDir/copySingleCellSignalsPeaksFiles.py --assembly mm10 --dry-run
 #   python3 $scriptDir/copySingleCellSignalsPeaksFiles.py --assembly mm10
 
 ##############################################################################
 # 3. Generate the trackDb .ra
 ##############################################################################
 # makeSingleCellSignalsPeaksRa.py reads the hub's mm10 stanzas, keeps the
 # cellBrowserMm10 subtracks, renames the composite to singleCellSignalsPeaks,
 # repoints every bigDataUrl at the local /gbdb copy, and writes the .ra with
 # group=regulation (ATAC signal/peaks sit with the ENCODE regulatory tracks).
 # The generator also: drops any subtrack whose source path is under a deprecated
 # "*.old/" dir, sets every subtrack "off" by default (users pick tracks from the
 # faceted selector), and assigns "priority" by cell class so same-class tracks
 # group together in the display. Labels/colors/facets are already resolved by
 # build_stanzas (see section 4); the generator carries them through unchanged.
 #
 #   python3 $scriptDir/makeSingleCellSignalsPeaksRa.py --assembly mm10
 #
 # https://github.com/ucscGenomeBrowser/kent/tree/master/src/hg/makeDb/scripts/singleCellSignalsPeaks
 
 ##############################################################################
 # 4. Labels, colors, and facets  (build_stanzas.py)
 ##############################################################################
 # build_stanzas.py derives the display metadata rather than copying the source
 # hub's cryptic cluster labels. The same logic runs for hg38 (see
 # doc/hg38/singleCellSignalsPeaks.txt).
 #
 # Cell type: the source cluster label is cleaned and, for datasets with coded
 #   cluster names, mapped to a canonical name via a paper-curated crosswalk
 #   (celltype-crosswalks/<collection>.tsv; codes decoded from Li 2021 PMID 34616068,
 #   Zhang 2022 PMID 36207411, Zhu 2021 PMID 33589836, and the Allen whole-brain
 #   taxonomy). Redundant synonyms are merged, QC clusters (doublet / low-quality /
 #   batch) are dropped, and commas become ';' (the faceted UI tokenizes cell
 #   values on commas). The full cell type is a searchable table column (_Cell_type),
 #   not a facet -- there are too many values.
 #
-# Cell class: each cell type is assigned one of ~22 broad classes
+# Cell class: each cell type is assigned one of 23 broad classes
 #   (celltype-crosswalks/celltype-class.tsv, built by build_celltype_crosswalks.py).
 #   The class drives BOTH the track color (one colorblind-conscious palette,
 #   celltype-palette.tsv, shared by hg38 and mm10 so a class is the same color on
 #   both assemblies) and the subtrack priority. Cell class is the primary cell facet.
 #
+#   "Other glia" (88,160,88) was added because "Other" was doing four unrelated
+#   jobs. Of 40 tracks corpus-wide sitting in it: 14 real glia with nowhere else to
+#   go, 14 peripheral glia deliberately placed there (Schwann / enteric -- left
+#   alone), 5 that are not cell types at all (All cell types / All cells /
+#   Neuronal and glial -- aggregate peak calls, a separate defect NOT addressed),
+#   and 6 genuinely other. The new class takes the glial aggregate labels (Glia,
+#   Glial), olfactory ensheathing cells, and the five fly glia subtypes; Schwann and
+#   enteric glia stay under "Other". Named for symmetry with the existing
+#   "Other neuron". The color came from a grid search maximizing minimum dE against
+#   the 22 existing entries under normal, deuteranope and protanope simulation,
+#   restricted to the widest free hue band; nearest neighbours are Epithelial
+#   (dE 17) and Astrocyte (dE 20), both above the palette's own closest pair
+#   (Other vs Unknown, 12.7). Before this, Glia and Non-Neuronal were both grey.
+#
 # Tissue / Life stage / Condition: filled per collection from hub_config.json
 #   (tissue_from_path, collection_lifestage, collection_condition) plus dataset-
 #   specific parsing in build_stanzas.py. SEA-AD, for example, gets its region
 #   (seaad_MTG/PFC -> middle temporal gyrus / dorsolateral prefrontal cortex),
 #   Aged life stage (donors 65-102), and ADNC neuropathology level (no AD / low /
 #   intermediate / high) from the path + "ADNC<n>" filename prefix; regions and
 #   ADNC categories per Gabitto 2024 (PMID 39402379) and Hawrylycz 2024 (PMID
 #   39402332).
 #
 # longLabel: rebuilt from the harmonized facets as
 #     <cell type>[, <condition if not healthy>][, <tissue if a specific region>]
 #     [, <extra>][, <variant>] (<dataset>)
 #   so the cryptic source short labels decode, e.g. "ADNC0 Astro" ->
 #   "Astrocyte, ADNC 0 (no AD), middle temporal gyrus (SEA-AD Brain ATAC)", and
 #   "LM.Macg.03" -> "Macrophage, muscle, 3 months (CATLAS Mouse Aging Brain)". The
 #   CATLAS aging age-in-months (the ".NN" filename suffix) is the <extra> descriptor
 #   -- kept in the label, not made a facet, so Life stage stays coarse (Adult/Aged)
 #   and consistent across datasets.
 #
 # Variant descriptors (track_variant): the cell-type harmonization above drops
 #   detail on purpose to keep the facets compact -- collapse_celltype strips trailing
 #   cluster numbers, rollup_celltype strips parenthetical qualifiers, and
 #   celltype_strip_suffix_regex strips peak-method suffixes. That detail is real, and
 #   without it hundreds of genuinely different subtracks collapsed onto one label
 #   (before this was added, hg38 had 925 subtracks sharing only 401 distinct
 #   longLabels, mm10 629 sharing 359). The variant descriptors put the meaningful
 #   distinctions back: how the peaks were called, which grouping level the track
 #   belongs to, which cohort, and signal vs peaks. They compose, since a peak file
 #   from one cohort needs both markers.
 #
 # Cluster identity (disambiguate_labels): any label still shared after the variant
 #   descriptors is qualified with its source cluster code, which is unique and is
 #   what the source paper calls the cluster. CATLAS mouse brain alone rolls
 #   MSGA1..MSGA13 into one "Medial septum GABAergic neuron", and ITL4GL1/ITL5GL2 into
 #   one "Cortical IT excitatory neuron" even though those encode different cortical
 #   layers -- right for a facet, wrong for a label. Every longLabel is now unique on
 #   both assemblies.
 #
 # shortLabel: rebuilt too (the source shortLabels were not in fact length-limited --
 #   they ran to 50 chars, carried underscores, ArchR filename tails and R-mangled
 #   names like X1126_HY_PAL_STR_Folh1_1). Built from the harmonized cell type,
 #   abbreviated word-by-word through a curated table (CT_ABBREV) only when it does
 #   not fit, plus the variant token and, where the longLabel distinction would
 #   otherwise be invisible, a compact facet token (SEA-AD region + ADNC level,
 #   CATLAS aging age: "Astrocyte MTG A0", "Basophil 3mo"). Budget is 22 chars: there
 #   is no hard limit in the tree (excluding this track, hg38's trackDb runs a median
 #   of 18 and 54% of tracks exceed the classic 17), but the left label area crowds
 #   past that and 22 leaves room for the distinguishing token.
 #
 # Spelling: known source misspellings are corrected before the plural/case merge so
 #   a typo'd variant lands in the same group as its correctly-spelled twin instead of
 #   becoming a second facet value ("Ventricular cardioyocyte" vs "Ventricular
 #   cardiomyocyte" were two separate Cell_type values). See CELLTYPE_SPELLING.
 #   The plural/case merge picks the singular form first and only then the most
 #   frequent -- frequency alone made the facet inconsistent, choosing the singular
 #   for 15 of 17 merged groups but the plural for Megakaryocytes/Oligodendrocytes.
 #
 #   IMPORTANT, if you add to CELLTYPE_SPELLING: the misspelled strings are load-bearing
 #   *keys* in the curated tables (celltype-class.tsv, and hub_config.json's
 #   celltype_rollup_overrides / celltype_expansions / celltype_expansions_global /
 #   celltype_tissue). Correcting a cell type without correcting those keys makes the
 #   lookup miss silently, and the track loses its rollup, tissue, broad class and color.
 #   That happened: correcting Broncial / Hematopoeitic / Syncitio / Glutaminergic, plus
 #   canonicalizing Megakaryocytes to the singular, quietly cost 16 hg38 tracks their
 #   Cell_class and color. The tables now normalize their keys on load (class_key,
 #   _ct_lower_key, _ct_matchkey) so either spelling matches, and any cell type with no
 #   broad class is reported in unclassified-celltypes.log plus counted in
 #   facet-coverage.md rather than silently becoming "unknown".
 #
 # Cell classes: Nephron progenitor used to come out as class "Neural progenitor". The
 #   decode tables carry a bare broad class of "Progenitor", and build_celltype_crosswalks
 #   mapped that to "Neural progenitor" -- right for the radial glia and neuroblasts that
 #   make up the rest of the class, wrong for kidney. Nephron progenitors are Six2+ cap
 #   mesenchyme, so they are now classed Stromal (NON_NEURAL_PROGENITOR); the HTML legend
 #   no longer lists them under Neural progenitor. Note celltype-class.tsv is hand-curated
 #   (the union of the mm10 crosswalk classes with paper-decodes/hg38_ct_class.tsv), NOT
 #   generated -- build_celltype_crosswalks.py does not write it, so a class change has to
 #   be made there too.
 #
 # One copy of the crosswalks: build_stanzas reads celltype-crosswalks/ out of the kent
 #   tree (XWALK_ROOT, overridable), not the copy in the hub-build dir. There used to be
 #   two independent copies and they drifted -- a stale local copy silently reverted the
 #   cell-class fix above while the kent copy looked correct.
 #
 # The curation of record is archived with the scripts:
 #   scripts/singleCellSignalsPeaks/build_celltype_crosswalks.py
 #   scripts/singleCellSignalsPeaks/celltype-crosswalks/       (per-collection crosswalks,
 #       celltype-palette.tsv, celltype-class.tsv, sea-ad-celltype-crosswalk.tsv)
 #   scripts/singleCellSignalsPeaks/celltype-crosswalks/paper-decodes/  (the paper-curated
 #       decode tables, each row's `note` giving its source justification)
 # build_celltype_crosswalks.py rebuilds the crosswalks + palette from paper-decodes/.
 # The general hub machinery that consumes them (build_manifest.py, build_stanzas.py,
 # build_hub.py, hub_config.json) lives in the cellBrowser repo, not in the kent tree --
 # it builds the whole super hub, not just this track:
 #   https://github.com/ucscGenomeBrowser/cellBrowser/tree/develop/ucsc/allTracksHub
 # build_stanzas finds the crosswalks archived here via XWALK_ROOT, so the hub and these
 # native tracks always color a cell class the same.
 
+##############################################################################
+# 4b. Two collections whose source hub stanzas are ignored
+##############################################################################
+# hub_config.json "collection_ignore_hub_stanzas" lists collections built as if they
+# had no curated hub at all: the cell type comes from the manifest original_track_name
+# (the source FILENAME) and the stanza is generated rather than copied. Normally the
+# wrangler-written hub stanza is the better source and is preferred everywhere --
+# corpus-wide its shortLabel classifies 1245 tracks that the filename cannot -- so this
+# is per-collection and must stay that way.
+#
+# Currently sea-ad-brain-atac and allen-basal-ganglia-atac. Both name their subtracks in
+# a DIFFERENT scheme than every piece of curation written for them, and both failed in
+# two independent ways:
+#
+#   Labels. SEA-AD files are ADNC0L2-3IT.bw -> original_track_name "ADNC0 L2-3IT", but
+#   its hub stanzas are named "L2/3IT_ADNC:0" -- the ADNC token moves from the front to
+#   the back and the layer name loses its hyphen. So label_sub (^ADNC(\d+)),
+#   celltype_leading_strip_regex (which strips a LEADING ADNC token) and the keys of
+#   sea-ad-celltype-crosswalk.tsv (L2-3IT, L5-6NP, Lamp5-Lhx6, Sst-Chodl) ALL miss, and
+#   72 hg38 cortical-layer tracks plus 89 mm10 allen-basal-ganglia tracks lost their
+#   broad cell class and their shared-palette color. Note the trailing " ADNC:0" is NOT
+#   what breaks it: lookup_class shortens by whitespace prefix, so "Lamp5 ADNC:0" ->
+#   "Lamp5" still resolves. The failures are the 9 spellings that differ between the two
+#   schemes (lookup_class("L2/3IT") is None; lookup_class("L2/3 IT") is Excitatory neuron).
+#
+#   Scaling. Both hubs declare a bare "type bigWig" with no data range, so 184 hg38 and
+#   89 mm10 tracks silently lost the default viewLimits bigWigInfo would have supplied.
+#
+# Add a collection here only when its hub stanzas are genuinely a different naming scheme
+# from its curation. The fix for a merely MISSING crosswalk row is to add the row.
+
+##############################################################################
+# 4c. Vertical scaling: autoScale group on the composite
+##############################################################################
+# The composite parent carries "autoScale group", and NO subtrack sets autoScale of its
+# own. Every selected subtrack then shares one vertical scale, so two tracks drawn at the
+# same locus can be compared directly; with per-track autoScale, tracks whose values
+# differ by orders of magnitude both drew full height and looked equally strong.
+#
+# It has to sit on the parent. hgTracks groups by tdb->parent (wigTrack.c setMinMax
+# accumulates each drawn track's limits into the shared parent tdb, and wigDrawPredraw
+# reads them back), and hubCheck errAborts on an individual bigWig that declares
+# "autoScale group" -- it "belongs in the parent composite stanza instead". 637 subtracks
+# had inherited exactly that verbatim from their source hubs, so build_stanzas now drops
+# autoScale via DROP_KEYS. A per-subtrack autoScale would also override the inherited
+# parent setting.
+#
+# Limits come from preDrawAutoScale scanning the preDraw array, i.e. the data in the
+# CURRENT WINDOW, not genome-wide -- so an outlying region elsewhere cannot flatten the
+# view. That is why grouping beats fixed per-dataset viewLimits, which was considered and
+# rejected: the spread of per-track maxima WITHIN one dataset reaches 20,117x (retina,
+# 0.80 to 16,092) and the maxima are outlier-dominated (a cortex-dev coverage track peaks
+# at 908,985), so a per-dataset ceiling would draw many real tracks flat.
+#
+# Grouping PER DATASET is not expressible here. autoScale group groups by tdb->parent and
+# a subtrack inside a view has the view as its parent, but a faceted composite cannot have
+# views: facetedCompositeUi walks tdb->subtracks exactly one level and builds <track>_sel
+# cart vars from the immediate children (hgTrackUi.c), so with a view level the leaf
+# bigWigs are never enumerated. None of the faceted composites in the tree use views.
+# Accepted cost: selecting across datasets with different normalizations (BrainVar
+# gene-activity maxing near 6 next to SEA-AD coverage near 72) draws the smaller tracks
+# small. The description page says comparisons are most meaningful within a dataset.
+#
+# Separately, a bigWig whose type line carries no data range gets one from bigWigInfo, so
+# the browser always has default viewLimits. Source hubs often set autoScale on the
+# COMPOSITE PARENT and this build flattens subtracks into one composite without copying
+# the parent, so the intent was lost in transit -- 232 subtracks corpus-wide had no range,
+# no autoScale and no viewLimits, and BrainVar (max ~6) drew as a flat line against
+# hgTracks' built-in 0:127.
+
 ##############################################################################
 # 5. Redundant allen-brain-science copies removed
 ##############################################################################
 # The Allen basal-ganglia dataset serves each unchanged per-cluster bigWig from all
 # four of its grouping directories (bg_regrouping_cl, bg_merge_D1_D2,
 # bg_merge_dorsal_ventral, bg_merge_D1_D2_dorsal_ventral). For 14 basenames those
 # four copies are byte-identical (md5-verified), so they rendered as four
 # indistinguishable subtracks and cost 4.4 GB of duplicated storage. build_stanzas
 # keeps the bg_regrouping_cl copy -- the cluster-level source the merges are built
 # from -- and skips the other three (allen_duplicate_skip; see
 # $CBHUB_OUT/allen-duplicates.log for the list). Where the four copies
 # genuinely differ (14 other basenames, different aggregations and different file
 # sizes) all four are kept and told apart by the grouping-level variant descriptor.
 # The 42 redundant files were deleted from the bed dir; the list is kept alongside
 # them in deleted-redundant-allen-copies.txt. Only the one directory-level /gbdb
 # symlink exists, so nothing needed cleaning up there.
 #
 #   xargs -a deleted-redundant-allen-copies.txt rm
 
 ##############################################################################
 # Counts
 ##############################################################################
 # 647 files resolved from the hub manifest (0 missing) and copied to the bed dir.
 # 42 byte-identical Allen copies were then removed, leaving 605 data files. 18
 # QC-cluster tracks (doublet / low-quality / batch) are dropped when the .ra is
 # generated, leaving 587 subtracks in the track (581 signal + 6 peak). Facet
 # metadata rows match the subtracks 1:1, and all 587 longLabels are unique.
+# Every subtrack has a broad cell class and a palette color (0 unclassified);
+# the 89 allen-basal-ganglia-atac tracks were the ones fixed by section 4b.