jbrowse-plugin-msaview 2.8.0 → 2.8.2

package/dist/utils/ncbiOrthologs.js

@@ -23,8 +23,14 @@ const DATASETS = 'https://api.ncbi.nlm.nih.gov/datasets/v2';
 const EUTILS = 'https://eutils.ncbi.nlm.nih.gov/entrez/eutils';
 // The species panel offered in the launch dialog, ordered from the reference
 // outward so a run that finds only close relatives still reads as a ladder.
-// Orthologs absent for a given gene are skipped rather than erroring, and the
-// index order here is the ROW order of the alignment (COMMON_TAX_RANK below).
+// Orthologs absent for a given gene are skipped rather than erroring.
+//
+// The index order is the order the sequences are SUBMITTED in
+// (`COMMON_TAX_RANK` below sorts `fetchOrthologGenes`' return), not the order the
+// rows are drawn in: the view lays rows out by the guide tree the aligner returns,
+// so a run on this list comes out grouped by relatedness rather than by this
+// list's own sequence. Reordering here changes what Clustal is handed, not the
+// picture.
 //
 // THE MAMMALS EARN THEIR PLACE, and the reason is measured rather than aesthetic.
 // The thirteen this list used to hold were one per major clade, which reads well

package/dist/version.d.ts

@@ -1 +1 @@
-export declare const version = "2.8.0";
+export declare const version = "2.8.2";

package/dist/version.js

@@ -1 +1 @@
-export const version = '2.8.0';
+export const version = '2.8.2';

package/package.json

@@ -1,5 +1,5 @@
 {
-  "version": "2.8.0",
+  "version": "2.8.2",
   "license": "MIT",
   "name": "jbrowse-plugin-msaview",
   "repository": {
@@ -51,7 +51,7 @@
     "puppeteer": "^25.3.0",
     "react": "^19.2.8",
     "react-dom": "^19.2.8",
-    "react-msaview": "^5.7.1",
+    "react-msaview": "^5.7.2",
     "rimraf": "^6.1.3",
     "rxjs": "^7.8.2",
     "serve": "^14.2.6",

package/src/LaunchMsaViewExtensionPoint/index.ts

@@ -1,3 +1,4 @@
+import type { OrthologParams } from '../MsaViewPanel/model'
 import type PluginManager from '@jbrowse/core/PluginManager'
 import type { AbstractSessionModel } from '@jbrowse/core/util'
 
@@ -21,6 +22,23 @@ interface LaunchMsaViewArgs {
   showBranchLen?: boolean
   querySeqName?: string
   highlightColumns?: number[]
+  /**
+   * Build the alignment from NCBI orthologs at launch time instead of naming a
+   * file: `{ taxId, geneCandidates }` is enough, and `taxa` and
+   * `proteinSequence` both default (see OrthologParams). This is the launch
+   * dialog's Orthologs tab, reachable from a session spec — so a link can say
+   * "NLRP1 across species" and the view builds it.
+   */
+  orthologParams?: OrthologParams
+  /**
+   * Hide any column gappier than this percentage, 100 being "hide nothing".
+   * A native react-msaview property, named here because it is the one setting
+   * that decides WHICH columns a freshly launched view opens on: proteins that
+   * differ in length put one row's private N-terminal extension at column 0,
+   * and everything else is gap there. Anything else react-msaview takes as a
+   * snapshot property passes through the same way.
+   */
+  allowedGappyness?: number
 }
 
 export default function LaunchMsaViewExtensionPointF(
@@ -40,15 +58,24 @@ export default function LaunchMsaViewExtensionPointF(
         ...rest
       } = args
 
-      if (!data && !msaFileLocation && !msaIndexedLocation) {
+      // `orthologParams` is a fourth source, and unlike the other three it names
+      // no alignment at all — the view builds one from NCBI at launch, which is
+      // the dialog's Orthologs tab reached declaratively.
+      if (
+        !data &&
+        !msaFileLocation &&
+        !msaIndexedLocation &&
+        !rest.orthologParams
+      ) {
         throw new Error(
-          'No MSA data or file location provided when launching MSA view',
+          'No MSA data, file location or orthologParams provided when launching MSA view',
         )
       }
 
       // inline data and the tree URL are native react-msaview snapshot props, set
-      // directly. Only sources needing launch-time resolution go through `init`:
-      // msaUrl (AlphaFold sniff) and the name-indexed bgzip block (no native loader).
+      // directly, and so is orthologParams (the model's own autorun picks it up).
+      // Only sources needing launch-time resolution go through `init`: msaUrl
+      // (AlphaFold sniff) and the name-indexed bgzip block (no native loader).
       session.addView('MsaView', {
         type: 'MsaView',
         ...rest,

package/src/MsaViewPanel/doLaunchOrthologs.ts

@@ -1,6 +1,7 @@
 import { cleanProteinSequence } from '../LaunchMsaView/util'
 import { launchMSA } from '../utils/msa'
 import {
+  COMMON_SPECIES,
   fetchOrthologRows,
   fetchProteinForGene,
   resolveGeneId,
@@ -31,7 +32,6 @@ export async function doLaunchOrthologs({
 }) {
   const { taxId, taxa, geneCandidates, msaAlgorithm, proteinSequence } =
     self.orthologParams!
-  const cleanedSeq = cleanProteinSequence(proteinSequence)
 
   const onProgress = (arg: string) => {
     self.setProgress(arg)
@@ -45,8 +45,29 @@ export async function doLaunchOrthologs({
     )
   }
 
-  // the query species is represented by the user's own transcript below
-  const wanted = new Set(taxa.filter(t => t !== taxId))
+  // The query row. The dialog always supplies it — it is the user's OWN
+  // selected transcript, which is what makes `connectedFeature` map genome
+  // coordinates through this row. A launch that has no transcript to translate
+  // (a session spec naming only a gene) falls back to NCBI's representative
+  // protein for the resolved gene, which is the same choice made for every
+  // other row, so the alignment is the one NCBI would build for that gene.
+  const representative = await fetchRepresentativeQueryProtein(resolved.geneId)
+  const cleanedSeq = proteinSequence
+    ? cleanProteinSequence(proteinSequence)
+    : representative?.sequence
+  if (!cleanedSeq) {
+    throw new Error(
+      `No query protein: none was supplied and NCBI returned no representative protein for gene ${resolved.geneId}.`,
+    )
+  }
+
+  // Every species the panel offers, when a launch names none. A spec that wants
+  // a narrower comparison says so; one that just wants "this gene across
+  // species" should not have to enumerate the list the dialog would have
+  // checked for it.
+  const wantedTaxa = taxa ?? COMMON_SPECIES.map(s => s.taxId as number)
+  // the query species is represented by the query row above
+  const wanted = new Set(wantedTaxa.filter(t => t !== taxId))
   const rows = await fetchOrthologRows({
     geneId: resolved.geneId,
     taxa: wanted,
@@ -54,7 +75,7 @@ export async function doLaunchOrthologs({
   })
 
   const treeMetadata: Record<string, Record<string, string>> = {
-    QUERY: await buildQueryMetadata(self, resolved.geneId, cleanedSeq),
+    QUERY: buildQueryMetadata(self, resolved.geneId, cleanedSeq, representative),
   }
   for (const row of rows) {
     treeMetadata[row.label] = buildRowMetadata(row)
@@ -76,34 +97,42 @@ export async function doLaunchOrthologs({
 }
 
 /**
- * The query row is the user's own translated transcript, so it carries an
- * Accession — which is what drives the automatic CDD overlay
- * (afterCreateAutoruns.autoLoadProteinDomains -> loadProteinDomains) — ONLY
- * when its sequence is byte-identical to the RefSeq protein that accession
+ * A failed lookup only costs the query row its domain overlay and, for a launch
+ * that supplied no sequence of its own, the alignment — so it is reported by
+ * returning nothing rather than by throwing here.
+ */
+async function fetchRepresentativeQueryProtein(geneId: string) {
+  try {
+    return await fetchProteinForGene(geneId)
+  } catch (e) {
+    console.warn('[msaview-orthologs] query protein lookup failed:', e)
+    return undefined
+  }
+}
+
+/**
+ * The query row carries an Accession — which is what drives the automatic CDD
+ * overlay (afterCreateAutoruns.autoLoadProteinDomains -> loadProteinDomains) —
+ * ONLY when its sequence is byte-identical to the RefSeq protein that accession
  * names. Attaching it unconditionally would put every domain box at an offset
  * whenever the user picked a non-representative isoform, which is a silently
- * wrong figure rather than a missing one.
+ * wrong figure rather than a missing one. A launch that took the representative
+ * protein as its query row passes that test by construction.
  */
-async function buildQueryMetadata(
+function buildQueryMetadata(
   self: JBrowsePluginMsaViewModel,
   geneId: string,
   proteinSequence: string,
-): Promise<Record<string, string>> {
+  representative: { accession: string; sequence: string } | undefined,
+): Record<string, string> {
   const transcript = self.orthologParams?.selectedTranscript
   const metadata: Record<string, string> = { 'Gene ID': geneId }
   const name = transcript?.get('name') ?? transcript?.get('id')
   if (name) {
     metadata.Transcript = name
   }
-  try {
-    const representative = await fetchProteinForGene(geneId)
-    if (representative?.sequence === proteinSequence) {
-      metadata.Accession = representative.accession
-    }
-  } catch (e) {
-    // a failed lookup only costs the query row its domain overlay, so it must
-    // not take down an alignment that is otherwise complete
-    console.warn('[msaview-orthologs] query protein lookup failed:', e)
+  if (representative?.sequence === proteinSequence) {
+    metadata.Accession = representative.accession
   }
   return metadata
 }

package/src/MsaViewPanel/model.ts

@@ -58,13 +58,24 @@ export interface BlastParams {
 export interface OrthologParams {
   /** NCBI taxon id of the assembly the query gene came from */
   taxId: number
-  /** taxon ids to include as rows (the query taxon is represented by QUERY) */
-  taxa: number[]
+  /**
+   * taxon ids to include as rows (the query taxon is represented by QUERY).
+   * Omitted means every species the launch dialog offers, which is what a
+   * launch that just wants "this gene across species" wants.
+   */
+  taxa?: number[]
   /** candidate gene identifiers off the feature, tried in order */
   geneCandidates: string[]
   msaAlgorithm: MsaAlgorithm
   selectedTranscript?: Feature
-  proteinSequence: string
+  /**
+   * The QUERY row. The launch dialog always supplies it, translated from the
+   * transcript the user picked, which is what `connectedFeature` maps genome
+   * coordinates through. Omitted — a session spec naming a gene and nothing
+   * else — the query row becomes NCBI's representative protein for the resolved
+   * gene, the same choice every other row makes.
+   */
+  proteinSequence?: string
 }
 
 /**

package/src/utils/ncbiOrthologs.ts

@@ -26,8 +26,14 @@ const EUTILS = 'https://eutils.ncbi.nlm.nih.gov/entrez/eutils'
 
 // The species panel offered in the launch dialog, ordered from the reference
 // outward so a run that finds only close relatives still reads as a ladder.
-// Orthologs absent for a given gene are skipped rather than erroring, and the
-// index order here is the ROW order of the alignment (COMMON_TAX_RANK below).
+// Orthologs absent for a given gene are skipped rather than erroring.
+//
+// The index order is the order the sequences are SUBMITTED in
+// (`COMMON_TAX_RANK` below sorts `fetchOrthologGenes`' return), not the order the
+// rows are drawn in: the view lays rows out by the guide tree the aligner returns,
+// so a run on this list comes out grouped by relatedness rather than by this
+// list's own sequence. Reordering here changes what Clustal is handed, not the
+// picture.
 //
 // THE MAMMALS EARN THEIR PLACE, and the reason is measured rather than aesthetic.
 // The thirteen this list used to hold were one per major clade, which reads well

package/src/version.ts

@@ -1 +1 @@
-export const version = '2.8.0'
+export const version = '2.8.2'