-
Notifications
You must be signed in to change notification settings - Fork 0
publication_plan
Created: 2026-05-23
mSystems (ASM) — first choice. Rationale: accepts software + database as co-equal contributions; broad environmental microbiology audience; no word limit; explicitly accepts metagenome case studies; FeGenie was published in Front. Microbiol. — stepping up in venue is appropriate.
Alternative: Nucleic Acids Research database issue (if framed as a database resource).
"In a metagenome from a redox interface — where iron oxidation and reduction coexist, where iron limits primary production, where metal resistance co-selects with iron cycling — which organisms are doing what, and with what genes?"
Do NOT frame as "FeGenie but better." FeGenie cannot answer this question. Efesto can. Lead with the ecological question.
Suggested title direction:
Efesto: an expanded HMM library and metagenome-aware pipeline for dissimilatory metal cycling annotation
| Section | Content | Status |
|---|---|---|
| Introduction | Gap: FeGenie cross-hits (MtrA/MtoA), no metagenome support, category errors, no metal resistance integration | Documented in curation MD |
| Library curation | Layer A/B methodology, category audit, 17 new models, MtrA/MtoA calibration | Fully documented |
| Software | Architecture: Pyrodigal-GV, operon detection (UniOP), TPM normalization, Anvi'o output | Code complete; needs architecture figure |
| Benchmark — precision/recall | Efesto vs FeGenie on reference genomes with ground truth | MISSING — critical |
| Benchmark — MtrA/MtoA | Cross-hit rate with FeGenie models; disambiguation with new models | Calibration report exists; needs tabulated FP comparison |
| Case study | ≥2 environmental metagenomes; show biological interpretation improvement | MISSING — critical |
| Performance | Runtime / memory across dataset sizes | Missing — simple to add |
Select 15–20 genomes with known metal cycling phenotype:
Iron reducers:
- Shewanella oneidensis MR-1
- Geobacter sulfurreducens PCA
- Desulfovibrio vulgaris Hildenborough
Iron oxidizers:
- Acidithiobacillus ferrooxidans ATCC 23270 (acidophilic — rusticyanin + Cyt579 test)
- Sideroxydans lithotrophicus ES-1 (MtoA canonical)
- Gallionella capsiferriformans ES-2 (MtoA canonical)
- Leptospirillum ferrooxidans (Cyt579 homolog, no rusticyanin)
Metal resistance positive controls:
- Cupriavidus metallidurans CH34 (multi-metal resistance)
Negative controls:
- Escherichia coli K-12
- Bacillus subtilis 168
Run FeGenie (original) and Efesto on same set. Compare:
- False positive rate — especially MtrA/MtoA cross-annotation
- Category correctness (iron transport miscalled as resistance)
- Recall on genes with known function
From existing hmm_library/_calibration/mtr_mto/calibration_report.tsv:
- N sequences misannotated by FeGenie model (scoring above 140)
- N correctly excluded by Efesto (below GA=580 / GA=520)
- Specific examples: MtrD family, betaproteobacteria DmsE
Already have the data — needs summarizing into a table (~1 day).
Two contrasting environments:
- Acid mine drainage (SRA has many — e.g., Río Tinto): rusticyanin/Cyt579 validation; expected high iron oxidation signal
- Iron-rich groundwater or freshwater sediment: MtrA vs MtoA separation matters most; iron stress markers (flavodoxin) expected if iron-limited fractions
Key story to find: one MAG/bin where FeGenie calls contradictory/ambiguous MtrA+MtoA and Efesto correctly resolves them by score + operon context. This is the concrete empirical validation of the cryptic cycling prediction (Díaz-González et al. 2025 mSystems).
Single script: run on 1, 10, 50, 100, 500 genomes. Report wall time + peak memory. ~0.5 day of work.
| Figure | Content |
|---|---|
| Fig 1 | Library composition — bar chart by category (active models, model sources) |
| Fig 2 | MtrA/MtoA score landscape — violin/dot plot across 3794-seq universe per class |
| Fig 3 | Software workflow diagram (input → gene calling → HMM search → operon → output) |
| Fig 4 | Benchmark precision/recall heatmap: Efesto vs FeGenie per category |
| Fig S1 | Category reclassification sankey or table |
Fill in reference genomes where each model fires (from benchmark step 1). Reviewers will ask; the column being empty is a liability.
Week 1 MtrA/MtoA summary table from calibration_report.tsv
Week 1-2 Reference genome benchmark (15-20 genomes, FeGenie comparison)
Week 2 Fig 1 (library stats) + Fig 2 (score landscape) + runtime benchmark
Week 3-6 Environmental case study (pick SRA datasets, run, interpret)
Week 6-7 Fig 3 (workflow) + Fig 4 (benchmark) + fill validated_in
Week 8-10 Write
FeGenie's own developers flagged this as unresolved. Zero other tools address it.
Evidence in hand:
- Cross-hits quantified: 1.9× / 2.1× the calibrated cutoff
- Calibrated against 3794-sequence universe
- Score landscapes: distinct classes separated
- TC/GA/NC values with biological rationale per boundary
Iron reduction vs iron oxidation is not a subtle ambiguity — it is a 180° functional difference. Any metagenome study using FeGenie models on mixed-redox environments carries this error.
| Tool | Iron cycling | Metal resistance | Metagenome workflow | Operon context |
|---|---|---|---|---|
| FeGenie 2020 | ✅ | ❌ | ❌ genome-designed | basic |
| MetHMMDB 2025 | ❌ | ✅ | ❌ | ❌ |
| Efesto | ✅ | ✅ | ✅ | ✅ UniOP |
No published tool spans all four. The combination matters: metal resistance without iron cycling context misses co-selection dynamics; iron cycling without metal resistance misses organisms that do both.
LaRoche et al. 1996 established flavodoxin as in situ iron stress marker in phytoplankton. No HMM pipeline has operationalized this for prokaryotic metagenomes.
Efesto has both models in iron_stress, correct category, outputs
per-sample counts. Enables a novel quantitative iron stress metric from
metagenomes. High value for oceanography, limnology, soil science.
FeGenie: Cyc2 only. Efesto: Cyc2 + Rusticyanin + Cyt579 → full downhill electron chain.
In At. ferrooxidans, rusticyanin is the most abundant periplasmic protein (~350 mg/mL). Annotating Cyc2 alone in AMD metagenomes underestimates acidophilic iron oxidation capacity. First library covering the complete diagnostic set.
SUF (sufB/sufC/sufS) vs ISC (IscS) ratio in metagenomes:
- SUF enrichment → organisms prepared for oxidative stress / iron limitation
- ISC dominance → housekeeping, iron-replete conditions
No existing tool separates these as distinct ecological signals. Application: iron-limited ocean gyres, AMD oxidative gradients, soil redox transitions.
FeGenie models: GitHub flat files, no provenance. Efesto registry: per-model source, nseq, cutoff basis, added_date, reference DOI, validated_in. FAIR data artifact citable independently of software. Increasingly required by journals (NAR, mSystems data availability standards).
| Flaw | Fix |
|---|---|
| No benchmark — all claims are assertions | Run reference genome set (task 1 above) |
| MtoA nseq=6 → low sensitivity | State explicitly as limitation; model will improve as Gallionellaceae MAGs accumulate |
| Calibration universe is title-searched (circular) | Supplement with manually searched MtrD sequences (already done); discuss limitation in methods |
| MetHMMDB is preprint (115 models) | Acknowledge as limitation; track if published before submission |
iron_acquisition typo (inherited from FeGenie) |
Fix in category names, file paths, outputs — do before benchmark so paths don't change |
| Flaw | Fix |
|---|---|
| Siderophore synthesis = 34% of active models (library imbalance) | Add narrative: siderophores are the numerically largest iron gene family in nature; imbalance reflects biology |
| No Archaea coverage | State as limitation; scope is bacteria |
| No Chlorobi / photoferrotrophic iron oxidation coverage | State as limitation; add FoxE/FoxY/FoxZ models in next version |
| Dissimilatory vs assimilatory iron reduction not distinguished | Add note in annotation guidance; future direction |
validated_in empty for all new models |
Fill from benchmark run |
| Source heterogeneity (4 sources, different training philosophies) | Document in methods; Layer A/B curation as harmonization step |
Find a MAG or enrichment metagenome bin where:
- FeGenie calls contradictory MtrA + MtoA (artifact of cross-hits)
- Efesto correctly assigns both by score margin + operon context
- The organism is demonstrably a dual-capacity iron cycler (cryptic cycling candidate)
This is what Díaz-González et al. 2025 (mSystems) predicted theoretically. Efesto can make it empirical.
Getting started
HMM library
Pipeline logic
Outputs and integration
Development