MAT discovery: Taphrinomycotina, Allomyces and unsearched-lineage probes - #353

Exploratory notes and probes for MAT discovery. No detector code change.

• Taphrinomycotina detector probe (33 genomes, job 32090545, commit 9c47592, preset v0.9.1): 7 of 8 Schizosaccharomyces genomes give mat1 calls (curated family, not independent). PM-family calls in 4 Protomyces genomes and Taphrina confusa.
• Genome-level matPi search (corrects earlier versions of this description): the BFD gene models lack the homeodomain gene of the PM locus, as does the curated T. deformans record. Searching the assemblies finds a homeodomain-like region on the same contig as an HMG-box gene, 7 to 18 kb away, in Protomyces inouyei, P. lactucae-debilis (2 assemblies), Taphrina confusa and T. communis, and at 40 kb in the T. deformans A2 control. 6 of 17 genomes against 0.71 expected under random placement (p 0.0005, floor of 2,000 draws; genomes not independent).
• Gene models and orthology (step 2 for Taphrinomycotina): exonerate models of the homeodomain gene; Pfam homeodomain ORF in each region. Reciprocal-best-hit test across genomes: the homeodomain gene is the same gene in all six genomes with a locus ORF (30 of 30 reciprocal, 32 to 53% identity); the HMG gene is the ortholog of the curated T. deformans matMc in the Protomyces assemblies and T. confusa (20 of 20 reciprocal; 44 to 45% in Protomyces, 69% in T. confusa). In T. communis and T. betulina the adjacent HMG gene is a different gene. Candidates for a curator: P. inouyei, P. lactucae-debilis, T. confusa. Expression, mating role and allele diversity are not shown. A bug in an ORF end coordinate was found and fixed on the way.
• Expression (RNA-seq): both genes of the pair are transcribed in T. deformans A2 (SRR8132818) and P. lactucae-debilis 12-1054 (SRR5576147). The unannotated homeodomain region is covered at every base at a depth at or above the median gene (A2 215x, 53rd percentile; 12-1054 1,432x, 80th percentile); the matMc-ortholog HMG genes are at the 83rd and 79th percentile. Read counts match ENA. One library per species, unstranded; the gene structure of the homeodomain gene is not resolved (A2 shows spliced reads downstream). Transcription is shown, a mating role is not.
• Transcript assembly (StringTie): in 12-1054 the homeodomain ORF is part of spliced multi-exon transcripts on the same strand; in A2 it lies on the strand opposite to the assembled transcript (unstranded run), so the strand and exons of the gene are unresolved there.
• Strand: the 12-1054 RNA-seq is reverse-stranded (measured over annotated genes); the homeodomain gene is on the minus strand (88.6% of read 1), the same strand as the HMG gene (method checked on that gene: 92.5%), 7.15 kb apart. The A2 run is unstranded, so A2 stays open. No stranded Taphrina data or long reads are public.
• The curated T. deformans matPi is the exact plus-strand translation of RHGF01000006.1:51,694-52,236 (181 aa, no stop, starts with Leu), the tblastn subject span: a region translation, not a gene model. It matches no BFD gene model because BFD did not annotate this gene. The RNA-seq shows the region is transcribed. No change was made to db/; replacing or relabelling the record is a curator decision.
• Allomyces flank-first idiomorph probe (3 genomes): candidates are not-conserved at the same rate as random genes (0.37 vs 0.37). Negative.
• Lineage survey and stage-1 scans with a chance test per lineage; Microsporidia homeodomain adjacency looks like a tandem family; one Paramicrosporidium HMG+homeodomain pair (single genome). Globomyces (2 genomes): conserved HMG cluster = species-specific tandem duplication.
• Annotation-free search of the 48 fragmentary assemblies (tblastn of HMG and homeodomain controls; unannotated_genome_search.py): 239 HMG regions in 18 genomes and 47 homeodomain regions in 21 genomes, and no HMG + homeodomain pair on one contig within 20 kb in any genome. Weak evidence: median N50 5.4 kb, so most contigs are shorter than the pair distance.
• Coverage: every target-phylum genome with BFD gene models scanned; 7 more assemblies scanned after Augustus gene prediction (weak evidence); 50 genomes remain unscanned (median N50 5.4 kb).
• Tests: new tests in 7 files pass. Full suite on commit 63e0d53 (job 32094325): 2038 passed, 8 skipped; later commits add a script, 3 tests and report text.

The curator merges.

🤖 Generated with Claude Code