Repository navigation
Releases
ParaScape 0.0.13.0
Full Changelog: v0.0.12.0...v0.0.13.0
ParaScape 0.0.12.0
Added
- Cryptosporidium has its first measured transcription slot.
Cp_transcription · oocystis filled from
GSE271211 (PMID 39354600): the three time-zero inoculum samples, undigested oocysts in triplicate, and not
the twelve digestion arms. It grades B with 2,566 of 3,923 genes (65.4 %). Genes the identifier map does
not reach stay missing rather than zero. The slot definition ships in this release; the Cryptosporidium
gene table is still built locally and is not part of the data download. - The rail down the left is a tree of what ParaScape can do. It was a flat list of six headings with the
modules of each under them. It now opens on a progression: under Start, the subject you have -- one
gene, a list of genes, a trait, a screen of your own, a question, a job -- and under each of those, the
modules that read it, ready ones first and a blocked one kept with the reason it cannot be taken. Under
every module sit the views behind it and the modules that follow it, so all 32 named views are reachable
from the rail. Every row is computed from the one module catalogue and the declared subjects
(parascape.tree, drawn byparascape.module_tree); nothing about what leads where is written down twice. - The navigator search finds a module by what it does and by what you would call it. Module names and the
names of the 32 views are ranked together, so "coexpression" opens on the tab of the Expression panel and
"expression" on the module. A module is also found by the subjects it reads, which come from its declared
ports, and by the words a parasitologist uses rather than the name on the tile -- "interactome",
"pulldown", "LOPIT", "rhoptry", "CRISPR", "UMAP", "AlphaFold", "RNA-seq", "gene set" -- none of which
matched anything before. The eight bands of the ranking are declared inparascape.modules. A search with
matches shows cards, as before; only a search with no match asks the assistant. - Five more modules can save their settings under a name. The Gene card, the Gene list card, Infer,
Predict and Screen each have a Recipes group, as Map and Network already did and as every spaCR screen does.
A recipe keeps the settings of a module and never the subject it is acting on, so a Gene card recipe says
how a gene is read and applies to whichever gene is in the card. The Predict recipe is the one that was
most obviously missing: a run is only comparable with another run made the same way. - Infer offers what to do next with an answer, as Predict already did. Both hand on a claim, and both
read the same list of what follows one.
Fixed
-
Six modules asked for a gene they do not read. Jobs offered "a gene id or a word of its product, then
Return" above the list of what is running, and Console, Gallery, Drop, Strategies and Scorecard did the
same. A module now offers a control for a subject only where it declares that it reads one, and the six
give the room to what they actually show instead. -
The name on a tile and the name on the module behind it are the same name. Five modules disagreed with
themselves: the tile said "Gene" and the module said "Gene card", the tile said Network was "a gene and its
neighbourhood, edges you define" and the module said "every gene and the links between them, yours in
focus". The masthead is read from the one catalogue, and where the module had the plainer words those are
now what the tile, the tree and the search say too. The Gene module is called Gene card throughout. -
A crowded row of buttons cut its labels at both ends. "Open a saved bundle…" was drawn as "pen a saved
bundle" and "Save everything as a bundle…" as "everything as a bur", so neither said which button it was. A
label too wide for its button is shortened with an ellipsis and given in full in the tooltip. -
Tables said what their columns were before anything was loaded into them. The Drop module said
"NTIFIER IN THE F" for "identifier in the file", the Scorecard said "NFERRED CLAIMS", and the Information
module cut "evidence (reference · dataset)" in half. -
The Gallery told you to "run walk hyperparameters on the Map tab". There has been no Map tab since the Map
module was built; it says to open the Map module and use "Find the best map for a label". -
The landing could vanish the moment the main window appeared, depending on whether a 400 ms timer fired
before the window manager reported its panels visible. The guard is a deadline read when the signal arrives
rather than a timer that arms, so there is nothing left to race. -
A result card drew a long name over its own edge: a strategy named by a whole sentence ran to four lines in
a square with room for three. -
The Toxoplasma bradyzoite slot was two thirds cat gut.
transcription · bradyzoite / tissue cystread
14 samples ofrna108740_Daythat the submitter declares enteroepithelial stages from Felis catus. They
now belong totranscription · in vivo enteric, as an alternative under policyone, so a log2 ratio is
never averaged with FPKM. -
The Toxoplasma tachyzoite protein slot was mostly not tachyzoites.
protein abundance · tachyzoite
read nine AP2XII-1-depleted abundance columns and three fold changes beside the untreated ones. It now
reads the untreated columns only, and the twelve have two slots of their own,protein abundance · pre-sexual reprogrammingandprotein change · pre-sexual reprogramming, kept apart. -
The Toxoplasma map is rebuilt on the corrected slots: 8,140 genes, 331 features (was 319).
Data
- Data release 0.0.12.0: 10,929 files, 944.3 MB, the same files as 0.0.11.0 with three replaced --
slots.json,slot_hierarchy.jsonandgraph.npz. An installed copy gets the slot corrections above
only from this download.
Changed
- The word "recipe" now means one thing: a saved bundle of module settings, as in spaCR. The way of asking one
biological question in the Infer module, which was also called a recipe, is now a question template.
parascape.recipesbecameparascape.question_templates(RecipeisQuestionTemplate,RecipeResultis
TemplateResult,RecipeStoreisTemplateStore,questions.as_recipeisas_template,
report.recipe_pdfisquestion_pdf), and the settings recipes took the plain name,parascape.recipes.
Files written by the old store have the same fields and still load. Code that still importsrun,close
or the other old names fromparascape.recipesfails with a message that names the new place.
ParaScape 0.0.11.0
Added
- Five further organism spaces, declared and installable as data packs: rat, cat, Anopheles gambiae,
Anopheles stephensi and Plasmodium vivax. Each carries its Ensembl gene build, and the rat and cat
spaces also Bgee expression calls and Ensembl Compara human orthologs. They are listed with their real
state and never shown as calibrated: a space that holds a gene build and no measurements can name genes
and hold data, and can answer no question until measurements arrive. - Acceptance checks for those spaces, with the measure chosen per species rather than applied blindly. The
95 % reviewed-proteome bar needs a reviewed proteome, and the less-studied species barely have one: rat has
8,239 reviewed proteins, cat 234, A. gambiae 303, but A. stephensi has 9 and P. vivax 11. Where the
bar would be a statement about nine proteins, coverage is checked instead against an independent gene list
that someone else's pipeline produced. - Leakage attacks for the new spaces, so a question answered in one organism cannot quietly borrow another
organism's labels, and a slot generator that refuses any column two organisms both claim. - Which labels may legitimately stand beside each other when a question is asked
(parascape.label_pairing): 228 of the 420 ordered label pairs are allowed, the rest refused by name. - Which strategy is applicable to which space, from the evidence that space actually holds
(parascape.strategy_applicability), rather than offering every method everywhere. - A multi-task strategy (cross-fitted auxiliary tasks appended as features), completing the five-rung
strategy ladder. - Frozen prospective localisation claims and a frozen filled-matrix manifest, so a later run can be held to
what was predicted before the answer was known. - Host transcriptome and proteome columns: monocyte and brain cell-type expression from the Human Protein
Atlas single-cell data, and human and mouse tissue proteomes from PaxDb v6.1 as parts per million of the
proteome, which is what makes two tissues comparable. - Plasmodium liver-stage, PfDBSP occupancy and mosquito-stage measurements, derived and shipped as their own
tables (see the fix below for why they are not merged into the node table). - A verdict sheet of 27 open source decisions, each with the evidence behind it, for Einar to rule on.
Fixed
- The mass-spectrometry search no longer runs out of memory. A timsTOF
.dacquisition is a directory, so
measuring it with one file size gave 0 bytes, the batch size was chosen as though the run were small, and
four large runs were searched at once; directories are now measured by their contents. The search cap is
24 GiB, because the dominant cost is the fragment index rather than the batch. - Every column of both shipped tables now resolves to an information category, and the loader refuses a
shipped table carrying a column that does not — an uncategorised column was previously silent. - The Plasmodium node table is held at the 168 columns the frozen functional recovery benchmarks were
measured on. Those runs chose 132 features out of that table; from a wider table the same method chooses
differently, so a wider table is a different experiment rather than the same one with more context. Four
new Plasmodium deposits are therefore derived and shipped as their own tables but held out of the node
table (deposits.HELD_OUT_OF_NODES) until the benchmark is re-prepared. - Removed a derived translation-efficiency ratio that had been computed and refused twice before, from two
instruments. The footprint count stays, because a count is a condition; the ratio does not. - The column picker reports how many columns were excluded to keep a question from answering itself, instead
of excluding them silently. - The Species menu lists every pack space with the state read from disk, rather than a written-in list.
- 144 of 160 candidate studies for empty slots are refused with the reason recorded, so a later pass does not
re-litigate them.
Data asset
parascape-data-0.0.11.0.tar.gz — 10,929 files, 950.5 MB. The Plasmodium table in it is the
168-column one the frozen functional benchmarks were measured on.
ParaScape 0.0.10.0
Data asset parascape-data-0.0.10.0.tar.gz (corrected Toxoplasma expression summaries). Package distributions are attached by the release workflow after PyPI publication.
ParaScape0.0.9.0 quantitative data
Original quantitative table components with source reconciliation, literal tokens, sample metadata, numerical QC and preparation records. Provisional data for ParaScape0.0.9.0; not a public software release.
ParaScape 0.0.9.0
Draft staging of original workbook data and complete binary-workbook processing evidence. Publication awaits complete data preparation, independent verification, installed application checks and release CI. No public release is claimed.
ParaScape 0.0.8.0
Draft data staging. Publication awaits the complete checked core data bundle, all hosted asset verification, installed application validation and exact-commit release checks. This draft is not a public release.
ParaScape 0.0.7.0
Expanded portable original measurements, submitted designs, reference identifiers, technical QC and source provenance. Dataset coverage remains partial; source quantities, identifier versions and biological independence retain their declared scope. Publication is gated on exact-commit CI.
ParaScape 0.0.6.0
Expanded portable original measurements, submitted designs, reference identifiers, technical QC and source provenance. Dataset coverage remains partial; source quantities, identifier versions and biological independence retain their declared scope. Publication is gated on exact-commit CI.
ParaScape 0.0.5.0
Expanded portable original measurements, submitted designs, reference identifiers, technical QC and source provenance. Dataset coverage remains partial; source quantities, identifier versions and biological independence retain their declared scope. Publication is gated on exact-commit CI.