A rigorous forensic reanalysis of 16 3'-tag RNA-seq libraries across four mycelial tissues: from overcoming sequencing vendor misalignment and uncovering unannotated rRNA giants, to eliminating depth confounders, resolving tissue specialization, and building the first draft genome-scale metabolic model.
When the sequencing provider delivered their initial analysis, the results were completely unusable. Here is how forensic sequence inspection overturned vendor errors and identified the true composition of the libraries.
The commercial provider aligned all 16 libraries to Saccharomyces cerevisiae (baker's yeast).
Gene IDs in the delivered matrix were SGD systematic names (YDL151C, YER174C, Q0045/COX1);
the 7,127 features matched SGD R64 exactly, and uniquely mapped reads were only 453–2,284 per sample (0.006–0.03% of input).
Delivered PCA, differential expression, and human-only MSigDB Hallmark gene set analyses were completely void.
Direct BLAST of the most abundant sequence read returned 100% identity to Pleurotus 28S ribosomal RNA. The reads themselves were intact; the NovaSeq X Plus 94 bp single-end libraries with 14 nt directional UMIs were high quality, but required an adapted pipeline grounded in the NASA GeneLab GL-DPPD-7101-G consensus specification.
Nuclear ribosomal DNA accounts for 19.9% to 69.7% of reads (mean 53.5%) across the libraries. Standard reference annotations collapse or omit these repetitive arrays, scattering reads into multi-mapping noise.
Of the reads surviving nuclear depletion, ~60.9% mapped to a single mitochondrial locus (CM148777.1:52,604–56,069).
This locus was completely unannotated in all public P. ostreatus GTF files, causing standard biotype-filtering pipelines to silently discard it as unassigned noise.
Full accounting of raw reads, ribosomal RNA, assigned mRNA, and library retention status after directional UMI deduplication.
| Well | Sample ID | Tissue | Raw Reads | rRNA Counts | mRNA Counts | mRNA Yield | Genes ≥ 10 | Quality Filter Status |
|---|
Rather than assuming the NCBI RefSeq representative was optimal, four candidate genome assemblies were systematically evaluated.
| Assembly Accession | Strain / Isolate | Assembly Level | Assembly Size | Unique Map % | Multi-Map % | Assignment Outcome |
|---|---|---|---|---|---|---|
GCA_056149245.1 |
BOM_ss5 (Selected) | Contig (16) | 41.4 Mb | 24.94% | 53.17% | Winner (+6.9% mRNA yield, resolves rDNA array) |
GCA_056149315.1 |
BOM_ss14 | Contig (16) | 43.5 Mb | 24.92% | 53.18% | Dikaryon partner to BOM_ss5 (differed by 0.02%) |
GCA_029852705.2 |
PC9.15 | Chromosome (13) | 35.3 Mb | 21.49% | 53.10% | Collapses nuclear rDNA into multi-mapping noise |
GCF_014466165.1 |
PC9 (RefSeq Reference) | Contig (16) | 34.9 Mb | 5.45% | 69.14% | Worst Performer (69% multi-mapping) |
The cultivar is P. ostreatus cv. "Harbor Blue P01", a commercial blue oyster mushroom.
BOM_ss5 is a single-spore isolate matching the cultivar. Crucially, BOM_ss5 uniquely resolves the tandem nuclear rDNA repeat array
(JBQVBD010000012.1:2,298,112-2,346,320) that other assemblies collapse, assigning +6.9% more mRNA counts across every single sample.
A prevailing hypothesis was that truncated annotated 3' UTRs caused high unassigned read rates in 3'-tag data. Extending 97.7% of gene models by up to 500 bp (neighbour-capped) increased read assignment from 15.5% to 15.6% (+0.1%). This rigorous diagnostic proved truncated UTRs were negligible; unannotated rRNA was the true culprit.
Four libraries yielded fewer than 15,000 non-rRNA counts, dominating PC1 with sequencing depth rather than tissue biology. Toggle below to observe how filtering noise libraries transforms the analysis.
Tissue specificity index τ computed on group means can easily manufacture false markers from single-replicate outliers. We applied a stringent cross-replicate robustness test: a marker is accepted only if every replicate of its tissue exceeds the maximum in all other tissues.
Nodules represent early fruiting-body primordia. Their signature is defined by protective and structural proteins:
AAD021_000251, τ = 0.984).AAD021_005235, τ = 0.978) and 17.8 kDa Class I HSP.AAD021_002063, τ = 0.941).The exudophore produces liquid guttation droplets. It exhibits a coordinated extracellular H2O2-generating and lignocellulose-degrading signature:
AAD021_011248, τ = 0.954).AAD021_002509, τ = 0.967).AAD021_003933).AAD021_003985).In initial mean-based analyses, Fuzzy Mycelium appeared to express ostreolysin A6 and dye-decolorizing peroxidase (DyP). However, cross-replicate auditing proved these signals were driven by a single outlier replicate (1F). Only 1 of 50 markers survived replicate testing (2%), and zero co-expression modules survived FDR correction. They are vegetative mycelium with no defensible differential signature in this dataset.
Genes where every replicate in the designated tissue exceeds the maximum of all other tissues.
| Tissue | Gene ID (BOM_ss5) | Annotated Protein / Description | Tissue Specificity (τ) |
|---|
The first genome-scale metabolic reconstruction (GEM) of Pleurotus ostreatus on defined Mycoponic Nutrient Medium (MNM v3), incorporating targeted backward-search MILP gapfilling and enzyme-constrained biomass flux.
A global MILP gapfill across the full ~31,000 ModelSEED reaction pool stalled indefinitely (>10.5 hours).
By implementing a targeted backward graph walk (depth 6) from each unproducible target, candidate reaction pools were restricted to 1.8k–8k reactions.
All three essential gaps (NAD: rxn07095, FAD: rxn00122, UDP-GlcNAc: rxn28025) solved cleanly in 2.5–9.3 seconds.
Biotin remained unreachable despite targeted gapfilling. Biological verification confirmed that many basidiomycetes are natural biotin auxotrophs. Because the mycoponic nutrient medium (MNM v3) contains malt extract and peptone supplying B-vitamins, biotin was properly added to the medium definition rather than forcing an artificial de novo synthetic route.
RNA-seq recovers a complete 73.2 kb mitochondrial genome that its own assembly failed to annotate, illuminating intron homing endonuclease dynamics, mitochondrial expression, and nuclear-mitochondrial retrograde signaling.
Public deposited annotations of the P. ostreatus mitochondrial contig (CM148777.1) contained almost zero functional features.
By combining RNA-seq read coverage clustering with tRNAscan-SE, Aragon, and protein homology, our corrected annotation recovered:
Across twelve Pleurotus mitogenomes, size varies widely from 60,694 bp to 102,950 bp. Uniform reannotation revealed that mitogenome expansion in the genus is driven almost entirely by intron and homing endonuclease gene acquisition:
Browse all 20 high-resolution publication and supplementary figures generated from this analysis. Click any figure to inspect it in full 250 DPI publication quality with contextual captions.
All five draft research manuscripts compiled directly from this codebase, with 100% audited and verified numbers.
Establishes the corrected reference, annotation, and computational pipeline. Unveils the silent failure of biotype rRNA filtering against unannotated organellar rRNA and tests 3' UTR extension limits.
Identifies the molecular mechanisms of the exudate-producing mycelial structure. Reveals robust alcohol oxidase (AOX1) and glyoxal oxidase (GLOX) H2O2 generation and early primordium nodule differentiation.
Presents the first genome-scale metabolic reconstruction (GEM) of P. ostreatus on defined mycoponic medium, featuring targeted MILP gapfilling, cofactor producibility, and secretome flux mapping.
Recovers the unannotated 73.2 kb mitochondrial genome using ordinary RNA-seq. Details 13 respiratory complexes, 25 tRNAs, Group I/II introns, and homing endonuclease expression.
Examines size dynamics (60–103 kb) across 12 Pleurotus isolates, proving that intron and homing endonuclease gene acquisition governs genome expansion (r = +0.96) alongside nuclear retrograde signaling.
Complete internal compilation documenting all 51 audit rules, methodology comparisons, reference genome benchmarks, and step-by-step pipeline reproducibility parameters.