v4.2.14

A symbol that names a type resolves to the type. "EPG neurons" — a class named by its symbol, "EPG" — matched a FAFB reconstruction's synonym first (that cell also carries "EPG" as a synonym) and answered with one neuron's 38 partners as what VFB knows about EPG neurons. A name of the shape symbol + category noun ("EPG neurons", "the EPG neuron", "Kenyon cell types") names a type, and no record's own label ends that way — so when such a name lands on an individual without matching its label exactly, the resolver now searches the symbol alone among classes and takes an exact match. EPG neurons now return the class (FBbt_00047030): term info, 214 images, 10 subclasses, 12 driver reports, and the ranked downstream partner table. A connectivity question no longer gets an unrelated subtype tree appended just for saying "cell types" (#59).

scRNA-seq expression columns now say what they are. Every column of a multi-cluster expression table read the same cell-type label, because that is all the tool showed; the cluster names themselves encode the difference — study, year, sex, tissue. Column headers now read Davie 2018 · whole fly, FCA 2022 · female head, AFCA 2023 D30 · male whole fly, with the cell type prefixed only when it varies across columns, duplicate headers numbered, and the table caption stating what the columns are. The clusters queried are also spread across datasets — one per study before a second from any — so a table compares studies rather than four sex/tissue splits of the same one (#58).

Unit suite 1,315; full 72-task battery, 72/72, 0 errors, run twice (once under CI, once locally at concurrency 4 against a production build). #60, feedback from Clare, 4 September.