Candidate mussel and clam studies to curate

A screened shortlist of real, public mussel and clam RNA-seq datasets, the backlog for WP5 of mussel_clam_integration_plan.md. It uses the same screen as the oyster backlog in candidate_studies.md. Every accession, run count and group structure below was read from ENA’s deposited run and sample metadata on 2026-10-12, not from paper text. Publications were matched through PubMed on the same date; a project with no publication listed had none that could be found.

This is a working backlog, not a commitment. Claim a study with aree start-study as for oysters (parallel_study_runs.md), and update this page when it is curated or rejected.

How the screen was done

  1. Every public RNA-Seq run in ENA for the six registered mussel and clam taxa was listed by project (scripts/list_ena_rnaseq_projects.py): 186 projects, from 8 for M. trossulus to 75 for Manila clam.
  2. Every title was read. About 40 had a resilience-relevant design and enough runs to consider.
  3. For those, the deposited design was recovered from sample attributes (scripts/inspect_ena_designs.py): groups and their sizes, the species recorded on each run, read depth, and layout.
  4. Publications were matched for the strongest candidates, to confirm what phenotype was measured and whether two projects share animals.

What decides the order

The oyster backlog’s criteria carry over: raw reads public, at least three biological replicates per group confirmed from run metadata, a measured resilience outcome preferred over exposure only, and RNA-seq first. Two more matter for these species:

  • Pooling needs the same species and the same phenotype. Evidence pools on species_taxid + phenotype + feature_type. M. edulis and M. galloprovincialis never pool with each other, and an acidification study scored on larval viability never pools with one scored as acidification tolerance. A study that would be the only one in its species-and-phenotype group is worth less now than one that completes a pair.
  • Two projects from one lab are not automatically two studies. Several candidates below come from the same group, sometimes from the same rearing program. Each pair is marked with what is known about whether the animals differ. Confirm from the methods before counting them as independent.

Tier 1: the first within-species pools

The acceptance target for WP5 is at least one within-species pool with k ≥ 2 and real standard errors for mussels and for clams. Three pools are within reach.

A. Manila clam, heat (k = 2) — done

Done 2026-10-17. Both projects are reanalyzed and on main: TUCCI2025_HEAT_PRIMING (PRJNA1132140, PR #41) and PERUZZA2023_HEATWAVE (PRJNA906528, PR #64; Peruzza et al. 2023, Mar Pollut Bull 193:115192, 10.1016/j.marpolbul.2023.115192). Both are QuantSeq 3′ Tag-seq (PRJNA906528’s ENA label says PolyA; the paper and the reads say otherwise), both are fitted on clams with tank_level_replication (4 tanks per arm in PRJNA906528), and PRJNA906528 fits sex as a covariate. Independence confirmed: same hatchery (SATMAR), but clams bought in February 2021 vs February 2022. Both pooled comparisons are exposure_only, so this pool replicates a heat response, not tolerance.

Claimed 2026-10-12 as TUCCI2025_HEAT_PRIMING (PRJNA1132140, draft PR #41), registered but not runnable yet. Reading the paper’s methods found three things the deposit does not show:

  • The libraries are 3′ Tag-Seq (Lexogen QuantSeq, 75 bp single-end), not full-length RNA-seq. The RNA-seq workflow’s length correction (Salmon effective length, tximport offsets in DESeq2) is wrong for 3′ counts, so the workflow needs a 3′ Tag-seq mode first. PRJNA906528 is from the same group at the same depth and layout; check its library type before claiming.
  • Tank is the unit of treatment (two 25 L tanks per condition) and tank membership is not deposited, so it cannot be fitted on tanks without the authors’ tank IDs.
  • The heatwave clams are survivors. RNA was taken 15 days after the heatwave, from the digestive gland, after about 25% of naive clams had died.
Project Runs used Design (from deposited metadata) Publication Size
PRJNA1132140 40 RNA-seq, single-end, about 7 M reads Primed vs naive × heatwave vs no heatwave, n = 10 per group Tucci et al. 2025, iScience 28:113108, 10.1016/j.isci.2025.113108 13.8 GB
PRJNA906528 40 RNA-seq, single-end, about 7 M reads 30 °C vs 20 °C for 30 days, digestive gland, n = 20 per group Peruzza et al. 2023, Mar Pollut Bull 193:115192, 10.1016/j.marpolbul.2023.115192 7.9 GB

Why first: the cheapest real pool in the screen (about 22 GB, small enough for an 8 GB workstation), large n, and a measured outcome. Tucci et al. report significantly higher survival in heat-primed clams under lethal heatwaves, so primed vs naive under a heatwave is a contrast between groups that differ in a measured survival outcome. That is the closest thing in this screen to the oyster backlog’s resistant-vs-susceptible lineages, though it is induced rather than heritable, and survival is a group property, not measured per sequenced animal. Register it as thermal_tolerance with the classification argued in the YAML.

Curation notes:

  • Both projects deposit 16S amplicon runs from the same animals under the same attributes (40 and 39 runs). Counting them doubles every group. Build the sample sheets with --library-strategy RNA-Seq (added for this case; the per-strategy breakdown is printed whenever strategies are mixed).
  • Pooling contrast. The comparison that pools across the two is heatwave vs control in naive clams (PRJNA1132140) with 30 °C vs 20 °C (PRJNA906528). Make those the meta_analysis_primary comparisons; primed vs naive under a heatwave is a second, unpooled comparison.
  • Sex is unbalanced in PRJNA906528 (20 °C: 15 male, 5 female; 30 °C: 10 and 10). Carry sex into the sheet and fit it as a covariate.
  • Independence. Both come from the University of Padova group (the priming project lists Milan, Peruzza and Bargelloni; the 2023 project is unpublished). The 2023 project is a 30-day non-lethal exposure deposited before the priming experiment was published, so separate animals are likely, but confirm sampling dates or source cohort before counting k = 2.
  • Tissue for PRJNA1132140 is not in the deposit; the paper gives digestive gland.
aree fetch-samplesheet --bioproject PRJNA1132140 --study TUCCI2025_RPHIL_PRIMING \
  --condition-attribute Treatment --condition-attribute "Thermal stress" \
  --attribute Clam_ID --library-strategy RNA-Seq
aree fetch-samplesheet --bioproject PRJNA906528 --study <STUDY_ID> \
  --condition-attribute temp --attribute sex --library-strategy RNA-Seq

B. Hard clam, salinity (k = 3)

Project Runs Design Publication Size
PRJNA975821 21 PE, 22 M reads Gill. Acute and chronic hypo- (10 psu) and hypersalinity (45 psu) vs 30 psu, n = 3 per group; aliases AL1/AL5 and AH1/AH5 suggest two acute time points Zhou et al. 2023, Mar Environ Res 192:106198, 10.1016/j.marenvres.2023.106198 52.5 GB
PRJNA1159784 9 PE Gill. Three months at 15 or 35 ppt vs 25 ppt, n = 3 Zeng et al. 2024, Fish Shellfish Immunol 154:109888, 10.1016/j.fsi.2024.109888 5.2 GB
PRJNA1288005 15 PE, 44 M reads Hemocytes. 72 h at 5, 15, 25 (control), 35, 45 ppt, n = 3 Zeng et al. 2025, Fish Shellfish Immunol 166:110609, 10.1016/j.fsi.2025.110609 102 GB

Three independent experiments from two labs (IOCAS Qingdao; University of Florida), two of them on gill. All are exposure_only for transcription: Zeng et al. 2025 measured survival (none died at 5 to 35 ppt; all died at 45 ppt by day 7) but per group, not per sequenced animal. Prespecify one hyposalinity comparison per study as meta_analysis_primary (for example chronic 10 psu vs 30 psu; 15 vs 25 ppt; 15 vs 25 ppt) so the pool compares like with like. Only the needed runs of PRJNA1288005 and PRJNA975821 have to be staged. PRJNA1159784’s run report carries no read counts; confirm depth after download.

C. Mytilus edulis, heat (k = 2, up to 4)

Project Runs Design Publication Size
PRJEB48959 35 PE, 51 M reads Gill. 7 salinity × temperature groups, n = 5 (ArrayExpress E-MTAB-11178). Control 23 ‰ at 5 °C; 30 and 33 °C at 23 ‰ and 15 ‰; 5 ‰ at 5 °C Barrett et al. 2022, Genes 13:155, 10.3390/genes13010155 109 GB
PRJEB41447 24 PE, 48 M reads Outer-fjord animals warmed to 22 °C (n = 5) or 32 °C (n = 4) vs unwarmed (n = 5), plus field collections on a warm day (E-MTAB-9786) Clark et al. 2021, Sci Total Environ 767:144366, 10.1016/j.scitotenv.2020.144366 85 GB
PRJNA1050756 58 PE, 56 M reads Two Atlantic Canada sources (Sober Island, NS; St. Peters Bay, PEI) × 20, 25, 30 °C; 30 °C has only 11 animals in total none found 319 GB
PRJNA1377038 118 PE, 63 M reads Diploid and triploid, slow and fast growers, 18 vs 26 °C none found 654 GB

The two Greenland studies form the pool: same research group (British Antarctic Survey and Aarhus), separate published experiments, comparable contrasts (32 or 33 °C vs a cold control). Stage only the contrast runs (about 15 libraries from PRJEB48959, 9 from PRJEB41447).

Curation notes:

  • Species identity. None of these deposits reports genotyping the animals, and M. trossulus co-occurs with M. edulis in Atlantic Canada. Every M. edulis study needs species_identity and, short of genotyping, species_assignment_unverified (validation enforces this).
  • PRJNA1050756: 11 animals at 30 °C against 23 and 24 at the lower temperatures suggests mortality at 30 °C. If so, the 30 °C transcriptomes are survivors, which biases that arm. Check before using it.
  • PRJNA1377038 is the only M. edulis candidate with a growth phenotype (fast vs slow growers) under heat, which makes it the most resilience-relevant. At 654 GB it is a Tier 1 study to run on the larger workstation, not first.

Tier 2: good, with a specific caveat

Project Species Design Context Caveat
PRJNA776346 M. galloprovincialis Larvae, pH 7.4 vs 8.1, static and at mid-trochophore, 3 replicate cultures Acidification; larval abnormality measured (Kapsenberg et al. 2018, described in the deposit) Cultures are pools of embryos from nine parental pairs, so n = 3 cultures. Alone in its M. galloprovincialis larval group, but the natural cross-species partner for the oyster larval OA studies once orthology exists (WP6).
PRJNA954971 M. galloprovincialis 30 vs 18 °C, gill and mantle, n = 3 Heat Clean but small, and the only M. galloprovincialis heat study with n ≥ 3, so k = 1 for now.
PRJNA611904 M. galloprovincialis Carbamazepine × low pH, n = 5, single-end Acidification × pollutant Low pH vs control is usable; the pharmaceutical arms are a different stressor.
PRJNA688298 M. californianus Copper-exposed larvae sorted into normal vs abnormal, n = 3 pools per dose Pollutant; within-exposure tolerant vs sensitive Rare design (outcome measured on the sequenced animals), but a pollutant stressor and pooled larvae.
PRJNA738278 Manila clam Two populations × −1, 12, 28 °C, n = 3 Heat and cold, population contrast Small n per cell; would add a third Manila clam heat study.
PRJNA733332 Manila clam “Zebra 2” strain vs ordinary clams × ammonia, n = 3 Ammonia stress, strain contrast Selected-strain contrast is valuable; ammonia maps to pollutant_exposure.
PRJNA621633 Manila clam Air exposure, re-immersion, control, n = 3 emersion_desiccation The first use of that stressor term; groups recoverable only from sample aliases.
PRJNA1391265 Manila clam Chemical, thermal and mixed priming, 108 RNA-seq runs plus 108 amplicon Heat priming Design not in the sample attributes; recover it from the publication or submitter. 290 GB.
PRJNA1289209 Manila clam Eggs from broodstock primed at low pH vs ambient, 30 runs Acidification, parental priming Larval outcomes measured per the description; design fields not deposited.
PRJEB91723, PRJEB91724 M. edulis Noise exposure, then Vibrio splendidus, 24 runs each Pathogen challenge Infection arm n = 6 vs 18 controls across noise groups; two projects to reconcile.
PRJNA1149726 M. edulis Francisella halioticida low- vs high-virulence strain vs sterile broth, hemocytes, 24 h Pathogen challenge Libraries are pools of animals, 4 per strain.
PRJNA914271 M. edulis Hypoxia, low salinity, food × two sites × adult and seed, 112 runs Multi-stressor Design is in neither attributes nor aliases; 343 GB.
PRJNA748538 Manila clam Low salinity vs control at 12 h, n = 3 Salinity Aliases (C123, T456) suggest each library pools three animals.

Same program, not independent: hard clam acidification

PRJNA887747 (larvae to 10 months, acidified vs normal seawater, n = 4 per age), PRJNA900255 (extrapallial fluid and hemolymph after one year, n = 8 per tissue; Schwaner et al. 2022, Int J Mol Sci 23:16066, 10.3390/ijms232416066) and PRJNA909768 (mantle at 22 months, varying pCO2) all come from the Stony Brook long-term acidification program. They look like three sampling points of one rearing experiment, not three studies. Register them as one study with a prespecified primary comparison, not as k = 3, unless the methods show separate cohorts.

Rejected, and why

Project Reason
PRJNA824625 (M. galloprovincialis, salinity) Libraries are pools per tank, two tanks per arm: the unit of replication is the tank, n = 2.
PRJNA770496 (M. galloprovincialis, heat) One library per temperature.
PRJNA995578 (mussel and oyster mantle, acidification) 12 runs per species deposited under one BioSample each: the runs are not shown to be distinct animals.
PRJNA1285511, PRJNA933718 (Manila clam, heat) 21 runs under one BioSample; 6 runs under two. Runs are not animals.
PRJNA478917 (Manila clam, hypoxia) One library per condition.
PRJNA764366, PRJNA685920 (hard clam, heat × hypoxia; air exposure) One library per condition or time point.
PRJNA339676 (Manila clam, low salinity) Two libraries per group.
PRJNA1013758 (Manila and grooved carpet clam, heat × habitat) 11 Manila clam libraries over four groups; the other species is not registered.
PRJNA560413, PRJNA249058 Population and phylogenomic sampling, no treatment.
PRJNA1227433 (mussel, cockle, grooved carpet clam × Vibrio) Three vs three per species at 11 M reads; kept in mind for WP6, not for pooling.

Not RNA-seq evidence, noted for later

  • Bivalve transmissible neoplasia: PRJNA749800 (M. trossulus BTN, 140 GB), PRJNA1249915 (hard clam neoplasia). disease_associated only; see the plan.
  • Byssal gland surveys (PRJNA1225934, M. edulis) and the multi-tissue atlases are annotation resources, not contrasts.
  • M. californianus has the only sizeable mussel bisulfite deposits (94 runs, two projects); they are the first mussel methylation candidates once a mussel RNA-seq study is through.

Sourcing notes

  • ENA was queried by tax_tree(<taxid>) with library_strategy="RNA-Seq". Mislabelled strategies are possible in both directions (see LUTIER2022_OA_TIPPING), so a project missing here may still hold RNA-seq.
  • The species on every run was read from ENA, not assumed from the title: PRJNA588016 (Gulf of Maine) mixes 20 M. edulis and 4 M. trossulus runs, and PRJNA743680 (hard clam breeding) is half Mercenaria campechiensis.
  • No ENA study record for these candidates carries a PubMed link; the publications above were found by searching PubMed for the design.