Hop atlas home · Overview · Changelog · Architecture · Adding_Data · Introgression · Ribbon · Dosage · Phylogeny · Bundle · Ingest · Ingest_Gbs · Ingest_Wgs · Ingest_Grasdi · Ingest_Rnaseq · Ingest_Assembly

Rendered from hop_atlas/ingest/WGS.md (2026-10-13)

Ingesting short-read WGS into the hop atlas web app (:8900): runbook reconstructed from history

Written: 2026-10-08. Status: reconstruction of what was actually run. No command below is new; every one is quoted from, or points at, a dated record. Nothing here is a tested end-to-end pipeline (see ../ADDING_DATA.md). Labels used: - GENERIC: parameterised, reusable as is (config/env/CLI argument). - COPY-AND-EDIT: a dated, cohort-specific script; copy it to your own dated folder and edit the named constants. - RECORD: documentation only ("not meant to be re-executed blindly"); read it, then write your own job. - (inferred): my reading of the records, not stated in them.

Path shorthands (same as annevo_eval/docs/INTROGRESSION_ATLAS_TECHNICAL.md, "TECH" below): MC = sync/2026-09-30/multires_clade, MCW = /data/hop_data/multires_clade, CI = sync/2026-09-29/clade_ab_imputation, CIW = /data/hop_data/clade_ab_imputation, PCD = sync/2026-09-25/per_chromosome_dosage_multiref, P = s3://general-purpose/skothenhill/archive/per_chromosome_dosage_multiref (PBSS), LR1 = sync/2026-10-06/lineage_refinement, LR2 = sync/2026-10-07/lineage_refinement_v2, L = /data/hop_data/lineage_refinement, SX = sync/2026-09-30/sex_chromosome_lineages, PP = sync/2026-09-30/phased_parents_switch_concordance, KR = sync/2026-09-30/cordifolius_KR, CAR = sync/2026-10-08/carey2026_hop_wgs, LL = cluster/lepton_linear (repo root, not annevo_eval/cluster/). REF = /data/hop_data/ref_21110M_male/male_ref.fa (21110M, Akagi numbering).

Untracked in git (2026-10-08 check): LL/, MC/, KR/, PCD/, CAR/. Only LR2/PIPELINE.sh and hop_atlas/tools/ are tracked.

1. Scope: WGS data that has entered (or tried to enter) the atlases

No WGS exists in our records for: Havill parents (GBS, 2 libraries each), divpanel / USDA / OSU panels (GBS), Saaz × HL1012 (GRAS-Di). The "comparative WGS" set (PMR-480/MN-101 etc.) was deleted 2026-10-02 (DATA_CATALOG comparative_wgs).

2. Ordered stages

S0. Decide the cohort shape first

S1. Download (ENA/SRA → PBSS = system of record)

Records (all RECORD / COPY-AND-EDIT; no generic downloader exists):

Where reads live now (PBSS): Apollo across two prefixes archive/PRJEB63565/reads/ + archive/hop_mapping_PRJEB63565_data2/, plus reads_repaired/ (ASSAYS 8241: take URIs from LL/sheets/apollo_reads_manifest.tsv, never a guessed prefix); archive/cascade_PRJEB63534_reads/, archive/wye_PRJNA1357051_reads/, archive/cordifolius_KR_reads/, archive/SW_reads/.

QC gate (required, from the gotchas): 1. Record count per mate == archive read_count (ASSAYS 3217). Bytes are a staging check only; ENA fastq_bytes ≠ staged size by 6–8 % (ASSAYS 5489). 2. For any .bz2→.gz rebuild, assert first header is @<ACC>. (ASSAYS 3371). 3. Repaired file must supersede the original in the manifest the drivers stage from (ASSAYS 5830; memory apollo-cross-truncated-runs). 4. Decode the true instrument from read headers (python3 annevo_eval/data_inventory/query.py --platform; Carey: CAR/scripts/build_sample_table.py prefix table). 5. Old Illumina (GA-II/IIx): check quality encoding; convert Phred+64 with KR/scripts/k06b_trim_requal.py (ASSAYS 8315).

Gotchas: ENA/ODP throughput from the annevo box is 0.02–0.6 MB/s; on Lepton ENA ~22 MB/s with ≤ 16 connections, ODP ~57 MB/s (ASSAYS 8491, 8211). EOS: transfer only via PBSS, never ENA on the cluster (CLAUDE.md). ENA publishes no read_count for ERR11641723 / ERR11641819 (ANCHOR_EXEMPT).

After download: annevo_eval/data_inventory/register_product.sh, entry in catalog.yaml, refresh.sh (CLAUDE.md; Carey has a catalog entry carey2026_wgs_reads).

S2. Read QC

• Tool + flags (identical in every WGS record): fastp --detect_adapter_for_pe --length_required 30 --thread N --json <acc>.fastp.json (LL/linear_job.sh trim_one, FASTP_MIN_LEN=30; KR/scripts/k01c_fastp_kr.sh adds -w 4 -z 3; CAR/PIPELINE.sh §2).
• Mandatory before pbrun: pbrun fq2bam 4.7.0-1 crashes with "illegal memory access" on untrimmed 250 bp PE (Wye parents) and raw 2×75 GA-IIx (KR); fastp first fixed both (ASSAYS 8261, 8307). One Wye parent on Apollo h2 additionally needed cropping fastp -b 150 -B 150 -A -G -Q -L (PCD/PIPELINE.sh).
• Gate: fastp before_filtering.total_reads == 2 × archive read_count (Carey; ASSAYS 5489 "the real completeness gate"). fastp aborting mid-file is how truncation was finally caught (memory apollo-cross-truncated-runs).
• GENERIC inside LL/linear_job.sh (DO_FASTP=1 default). No standalone QC tool.

S3. Alignment

Reference: 21110M male_ref.fa for everything the introgression atlas uses (Chr01–09, ChrX, ChrY; Chr01–05 + ChrX reverse-complemented vs drHumLupu-forward; TECH §1.1). Dosage additionally uses apollo_h1, apollo_h2, saaz_h1, hl1012_h1, cascade (PCD/PIPELINE.sh §0). Carey also aligned to drHumLupu (and MN1421/MN586 HAP1 for 8 wild runs, QC only).

Tool: GPU pbrun fq2bam (Parabricks nvcr.io/nvidia/clara/clara-parabricks:4.7.0-1), standing rule (memory use-parabricks-for-alignment). Needs the classic bwa index (.amb .ann .bwt .pac .sa), asserted .ann length == sum of .fai (ASSAYS 4.2f; LL/upload_refs.sh).

WGS duplicate rule (inverted vs GBS): markdup ON (omit --no-markdups; add --out-duplicate-metrics); downstream exclude dups (--ff UNMAP,SECONDARY,QCFAIL,DUP; depth -F 0xF04). Gate: samtools view -c -f 1024 <bam> > 0 for WGS, == 0 for GBS (LL/README.md "Five things" #2; ASSAYS 8271). Pooled invocations need a unique LB per sample, else cross-plant duplicates (ASSAYS 8281; job_ref.sh LB_CONTROL=1).

Read group: @RG\tID:<run>\tSM:<sample>\tPL:ILLUMINA\tPU:<run>\tLB:<sample or library>; PU is mandatory (memory use-parabricks-for-alignment). Multi-run samples (Pub_740M) = several --in-fq pairs in one call (align_sample in LL/linear_job.sh).

Which alignment fed what (this matters — they are different BAM sets):

QC gates used: primary records (samtools view -c -F 0x900) == fastp output reads (Carey; KR); vs archive read count (READCOUNT_MANIFEST, linear_job.sh; memory verification-must-be-able-to-fail: BAM == fastp is self-referential); samtools flagstat; dup gate; mosdepth -n --fast-mode -F 1796 for measured mean depth (Carey). Per-sample QC table: CAR/scripts/build_sample_table.py → CAR/tables/carey2026_sample_qc.tsv (best template).

GENERIC: LL/linear_job.sh + one config/<pop>_<ref>.env + sheets/<pop>_sample_sheet.tsv (run → alias) + upload_inputs.sh + submit_linear.py. COPY-AND-EDIT: PCD/scripts/lepton/* (cohort list and read sources in 00d_build_manifest.py), CAR/scripts/lepton/* (sample sheet, refs, bucket prefix), KR/scripts/k01*. Compute site: all WGS alignment on Lepton (or local A6000 for KR). Nothing in the WGS records ran on EOS (inferred from records found). Lepton→PBSS is degraded; move data via AWS s3://bionemofw + annevo_eval/data_inventory/aws_bridge.sh (CLAUDE.md; ASSAYS 4.2e).

S4. Calling / allele counting

The atlas does not use genotype calls. It uses per-allele read counts (FORMAT/AD) at fixed site catalogues from per-sample BAM pileups.

S4a. Introgression base (clade A/B) counts — route B. MC/scripts/m01_count_one.sh <cohort> <sample> <pbss_key> WGS [targets], driver MC/scripts/m01_run_jobs.sh <jobs.tsv> 8 [targets]:

bcftools(1.20) mpileup -f $REF -T <targets> --min-MQ 20 --min-BQ 20 --ff UNMAP,SECONDARY,QCFAIL,DUP -a FORMAT/AD -Ou <presigned BAM url> \
 | bcftools query -f '%CHROM\t%POS\t%REF\t%ALT[\t%AD]\n'   -> MCW/bamcounts/<cohort>/<sample>.tsv.gz + .done   (FAIL if <= 1000 lines)

• Targets: full catalogue CIW/diag/targets_all.tsv.gz (29.2 M sites; Cascade/Wye WGS parents) or the 500 bp thinned MCW/jobs/targets_apollo_500bp.tsv.gz (Apollo, 2.34 M sites, "to bound output size"). No script generates the job tables (MCW/jobs/cw_jobs.tsv, apollo_jobs.tsv) or targets_apollo_500bp.tsv.gz (TECH §1.2).
• Route K (single plant, KR): KR/scripts/k02_pileup_catalog.sh <sample>, same flags, per chromosome (-r <chr>), full catalogue.
• Implicit mpileup defaults (-d 250, BAQ on, anomalous pairs skipped) are not overridden (TECH issue I-21). Supplementary alignments still pass --ff UNMAP,SECONDARY,QCFAIL (ASSAYS 7924).
• GENERIC (m01 is CLI-driven), but hard-codes REF, BCF and output root.

S4b. Joint VCF (not used by the served atlas for WGS cohorts). LL/config/apollo_21110M_untargeted.env: ROLE=call UNTARGETED=1 CALL_FF=UNMAP,SECONDARY,QCFAIL,DUP MIN_MQ_SET="0 20", joint call of 186 BAMs, no site list, no depth floor → PBSS lepton_linear/apollo_21110M_untargeted/calls/apollo_21110M_cohort_joint_mq{0,20}.vcf.gz (in that prefix, both/union = the superseded targeted design; trust only *_untargeted markers — memory apollo-cross-coverage-design). Used for route-V baseline Apollo (CI/PIPELINE.sh:44-45, 3 × 100 kb windows per 5 Mb) — superseded by route B. Gate rule: test an MQ floor by INFO/MQ minimum and depth, never by "mq20 < mq0" record counts (ASSAYS 5634).

S4c. Lineage SNP-layer counts. LR1/scripts/l02_count_one.sh <bam> <sample> <outdir> WGS (WGS ⇒ targets_wgs500.tsv.gz from l01b_wgs_targets.py, --ff …,DUP); job lists l02_make_wgs_jobs.py → L/jobs/{kr_wgs,cascade_wgs,wye_wgs,apollo_wgs}.tsv; driver LR1/scripts/l02_run_wgs.sh. Then LR1/scripts/l03_build_sparse.py <cohort> → L/sparse/<cohort>/. Carey ran the same mpileup per chromosome inside the GPU job (CAR/PIPELINE.sh §2) → L/bamcounts/carey2026/.

S4d. Tag-space counts (subspecies layer). samtools view -F 0xD00 <bam> | kt --mode scan --out <s>.kt --keep-loci L/tagspace/catalogue/keep_loci.u64 (WGS; GBS uses -F 0x900 … --mode gbs5scan); kt built from LR2/scripts/kt.cpp; driver LR2/scripts/kt_run_wgs.sh → LR2/scripts/t03_tag_sparse.py <cohort> 8 / t03_run.sh. Carey: in-job → L/tags/carey2026/.

S4e. Sex-chromosome pileups. SX/scripts/s10_pileup_one.sh via s10_run_jobs.sh $W/jobs/s10_jobs.tsv 8: bcftools mpileup -f $REF -T targets_sex.tsv -r ChrY,ChrX:225000001-268239420 --min-MQ 20 --min-BQ 20 --ff UNMAP,SECONDARY,QCFAIL,DUP -a FORMAT/AD -Ou "<bam url>##idx##<bai url>" → /data/hop_data/sex_chromosome_lineages/pileup/<cohort>/<sample>.tsv.gz, then s13_build_counts.py. Carey: not run.

S4f. Dosage read counts. In-job count_one_v2.sh BAM SAMPLE COHORT REFTAG WGS OUTDIR (PCD/scripts/lepton/): one pass over samtools view -F 0x900; primary / unmapped / dup / MAPQ<20 / MAPQ≥20 per sequence; WGS bins 100 kb of MAPQ≥20, non-dup, non-QCfail read starts (bins.tsv.gz) + any-MAPQ (binsall); FAIL if 0 primary records. Pulled with PCD/scripts/10_run_local_analysis.sh into /data/hop_data/per_chromosome_dosage_multiref/counts/<ref>/<pop>/ (WGS from <ref>_wgsmd). Allele balance for WGS: Apollo only from the existing full-depth hap1 call set (PCD/scripts/02_apollo_h1_existing_ab.sh); at ~2× no AB exists (ASSAYS 8251). COPY-AND-EDIT.

S5. NA/EU base calls (introgression atlas)

COPY-AND-EDIT points for a new WGS cohort (route B): - m03 reads its sample list from CIW/cohorts/<coh>.npz (the route-V baseline) and asserts every sample has a .done; a cohort with no baseline needs the OSU fork pattern (/data/hop_data/osu_introgression_atlas/scripts/o03_build_bam_cohort.py, sample list from a text file). - m04: TRUTH, TRIP dicts (lines 43–57), CROSS_VCF; OSU fork o04_impute_multires.py adds cohort lines. - m05/m07: FINAL, parent map ("apollo": ("Apollo_mother","Pub_740M")), ASSAY label strings. - Mixed-assay call sets share one dA/dB (TECH I-20). Fit the EM on ≥ several samples; single plants → route K. QC gate: m05 truth accuracy (Apollo_mother is a truth plant vs Apollo h1/h2, scored on leave-plant-out sites), callability, parent-inconsistent rate; viewer warning thresholds callable < 0.90 / truth < 0.99 (TECH §4.5). KR has a row-specific test k21 (KR/scripts/viewer_tests/).

S6. Viewer export for the base atlas

No generator (TECH §10.2, §11.2). What was done: - Apollo: MC/scripts/m07_viewer_export.py → multires/atlas_{samples,native,transmission}_<res>.json; detail sync/2026-09-29/universal_introgression_atlas/scripts/u17_multires_detail.py → /data/hop_data/universal_introgression_atlas/viewer_detail_mr/ (TECH §10.1). - KR: KR/scripts/k08_viewer_export.py (group "cordifolius (KR) — WGS 75 bp", detail files, global similarity recomputed), tests KR/scripts/viewer_tests/run_tests.sh <res> + p08_test_summary.py. - New cohort pattern: OSU o00_labels.py → o07_viewer_export.py → o08_preset_patch.py → o17_detail_osu.py → o09_swap_into_served.sh, all under /data/hop_data/osu_introgression_atlas/scripts/; full 15-step sequence in TECH §10.2. COPY-AND-EDIT (parents, labels, PANEL_ORDER hard-coded). - Carey 2026 stopped here: no multires/atlas_samples_<res>.json entry, so the layer exports skip it (CAR/layers/README.md).

S7. Layers

Subspecies / lineage (hop_atlas/components/INTROGRESSION.md §4.3): 1. SNP layer: LR2/scripts/c01_call_levels_v6.py <cohort> --groups $GF --nproc 8 --tag _v6 → L/calls/<cohort>_v6/ (c01_run_all_v6.sh). Edit CALLSET and PARAMS dicts: WGS cohorts without their own fit borrow 'MCW/apollo_bam' (KR, Carey). 2. Tag layer: LR_ROOT=$L/tagspace python3 c01t_call_levels.py <cohort> --groups $GF --nproc 8 --minref 1 --tag _t4 --l0-from '$L/calls/{coh}_v6' → L/tagspace/calls/<cohort>_t4/. Takes sex and the ChrX L0 constraint from the served sexchrom JSON; cohorts absent from it are scored diploid on ChrX (Carey; CAR/layers/README.md). 3. Register: add the cohort to COH in e03_viewer_export.py, e04_detail_export.py, l23_lineage_v3_test.py, and to LIMITED_COH (WGS: European-lineage calls from GBS-trained groups are not shown; WGS parents inside GBS cohorts via LIMITED_KEYS). Carey did this as copies in CAR/layers/ used through ATLAS_LAYERS_EXPORT_DIR and ATLAS_LAYERS_CHECKS. 4. Export/test: hop_atlas/tools/atlas_layers.py lineage|test (S9). - Depth cap test for WGS lineage: LR2/scripts/x07_cap_test.sh, x08_cap_compare.py (cap 1/2/16).

Sex lanes (TECH §8; SX/PIPELINE.sh): s13_build_counts.py → s14_sex_table.py → s15_y_lineage.py (WGS plants in GBS cohorts and KR use the Apollo reference-female masks) → s16_x_caller.py --no-mask → s16b_hxr_male_mask.py → s16_x_caller.py (1 Mb X and 5 Mb yb MSY bins only for WGS plants) → s24_viewer_export.py (asserts exactly 1,272 samples; cannot regenerate the current file, TECH I-34) → s27_viewer_integrate.py. OSU extension pattern: o11_sexchrom_osu.py merges into the served JSON from a backup. Sex sources: annevo_eval/sample_identity/sample_sex_verified.tsv (Cascade WGS parents: "parent_library_X_test_not_applicable", sex = Y k-mers + role). KR sex = F by PI decision. COPY-AND-EDIT.

Phasing / concordance (PP/PIPELINE.sh): p01_phase_concordance.py <res> ×4 (hard-coded SRC, parent and ASM dicts, e.g. Apollo mother phase from Apollo_h1/h2 if genotype agreement ≥ 0.95), p03_bfreq_transmission.py, p05_validate_concordance.py, p04_export_viewer.sh; then ChrX sex-aware via hop_atlas/tools/atlas_layers.py xconc $STG (LR2/scripts/x11_sexaware_x_concordance.py). Crosses only. COPY-AND-EDIT (OSU fork o01_phase_osu.py).

Pedigree: atlas_layers.py pedigree $STG (GENERIC); name matching in LR2/scripts/p01_pedigree_build.py (SYN, UNVER, EXCL, CROSS); new cross = one CROSS line.

S8. Dosage atlas

Dependency order (hop_atlas/components/DOSAGE.md §2–3): S3 per reference (<ref>_wgsmd) → S4f counts → clade calls from S5 (build_cn.py reads MCW/calls/<CALLSRC[pop]>_1Mb_calls.tsv.gz; ancestry-aware CN needs them) → v1 assets (/data/hop_data/hop_atlas/dosage_sources/2026-09-28/dosage_explorer/scripts/build_all.py, build_assets.py <ref>) → v2 builder /data/hop_data/dosage_explorer_v2/stage/scripts/rebuild_and_validate.sh (build_all_cn.sh → worked_example.py → fill_defs.py → make_page.py → validate_cn.py) → hop_atlas/tools/relocate_dosage.py. - Hard-coded lists to edit: build_all_cn.sh (ref × pop loops), validate_cn.py and check_cn_browser.py (REFS, POPS), v1 build_all.py, mr_config.py (REFS), build_cn.py (CALLSRC, FRAME, REFHAP, EXCLUDE_BASE). - WGS specifics: 100 kb bins; duplicates excluded; Apollo depth from ~2× subsample; X not readable as copy number in Cascade/Wye/Apollo (memory per-chromosome-dosage). Pan-X instrument = annevo_eval/methods/sexchrom_dosage/hop_xdosage (count --input <bam>, call --manifest), not in the served atlas. - Generic package: annevo_eval/methods/autosomal_dosage/hop_adosage (cn --bins … --manifest … --assay …) — merged to main in commit 029bf81; DOSAGE.md still says "unmerged branch" (stale). - KR and Carey: not in the dosage atlas.

S9. Stage → build → test → deploy (GENERIC)

T=hop_atlas/tools/atlas_layers.py        # introgression layers (annevo_eval/scripts/atlas_layers.py is a 285-byte shim)
python3 $T stage --name <change>; STG=/data/hop_data/lineage_refinement/stages/<change>
python3 $T lineage $STG; python3 $T pedigree $STG; python3 $T xconc $STG
python3 $T serve $STG --port 8907; python3 $T test $STG --port 8907; python3 $T deploy $STG      # deploy needs PI approval
D=hop_atlas/tools/atlas_deploy.py        # dosage: stage dosage / build dosage / test dosage / deploy dosage [--dry-run]

Base-atlas JSON (S6) has no stage tool of its own; OSU used a hand-copied stage on :8901 and o09_swap_into_served.sh (md5-guarded). Rules: hop_atlas/README.md (never copy an old stage over the served copy; CHANGELOG entry; PI approval).

3. Per-cohort differences

4. Same plant, different assay (WGS ↔ GBS/other)

Canonical table: annevo_eval/sample_identity/{sample_identity.tsv,identity_pairs.tsv,README.md}. Built by sync/2026-10-03/atlas_evidence_rebuild/scripts/e01_identity_table.py + e03_lineages.py (never hand-edit; re-run sync/2026-10-04/verified_sex_update/scripts/v01_identity_sex.py after). Same-plant groups (P####) = connected components of merged=True edges; status ok / duplicate / unresolved / non-progeny / different-plant / mislabel / mixture.

Naming: each library stays its own atlas row, keyed <cohort>|<sample> (WGS keys: run accession for Cascade parents and Carey; alias for Apollo/Wye/KR). Rows are never merged; identity is a group label, plus pedigree matching (p01_pedigree_build.py: exact / variant / synonym / approximate / unverified).

Identity checks used (records):

Recorded caveats: KING thresholds calibrated on GBS–GBS pairs; GBS-vs-deep pairs fall below them (REPORT calibration note). Cascade WGS mother vs OSU Cascade GBS is unresolved in sample_identity.tsv. Carey overlaps are "recorded, not resolved": Carey Cascade SRR37629444 is a different library from ERR11608298; Carey 19058M possibly = USDA 19058_M (DATA_CATALOG carey2026_wgs_reads). The OSU cross uses WGS ERR11608298 as mother in osu2022005_21110M, flagged SUPERSEDED (mother = WGS) in the explorer (sync/2026-10-02/osu_explorer_update/README.md:25); the served OSU atlas uses the OSU GBS cascade library as mother (TECH §1.3).

5. Missing or ambiguous steps

1. No base-atlas generator (S6). Every WGS cohort since Apollo entered by a fork (KR k08, OSU o07). Carey is blocked here.
2. Route-B job tables and the Apollo 500 bp target list have no producer script (MCW/jobs/*.tsv, targets_apollo_500bp.tsv.gz).
3. Apollo atlas rows use the ~2× subsample BAMs, not the ~9× linear BAMs. No record says this was chosen for the atlas; it follows from MCW/jobs/apollo_jobs.tsv → m21_wgsmd + PCD/scripts/00d_build_manifest.py (inferred). Full-depth BAMs exist under PBSS lepton_linear/apollo_21110M/ (inferred path from LL PUSH_PREFIX defaults).
4. EM refit when adding samples to an existing cohort: undocumented (ADDING_DATA.md).
5. Sex lane for a new cohort: s24 asserts 1,272 samples; only the OSU merge pattern exists. Not run for Carey.
6. Dosage add-cohort: no procedure; ~6 hard-coded lists; WGS allele balance needs a full-depth call set.
7. Download/QC: no generic tool; integrity checks differ per cohort (Cascade parents: gzip -t only — record count vs ENA never recorded for them).
8. Stale/contradictory records: sample_identity.tsv row kr:KR lists run_accession DRR024404–407 (the SW75 runs), not KR's DRR024452–456. DOSAGE.md says hop_adosage is unmerged (merged in 029bf81). ADDING_DATA.md/INTROGRESSION.md cite cluster/lepton_linear correctly but the task brief's cluster/lepton_linear does not exist. TECH §10.2 says the OSU sync scripts folder is empty (it is not).
9. Instrument for Cascade WGS parents never decoded from headers in any record found.
10. Carey status: 15/22 runs FAILED/absent in the 2026-10-08 19:07 QC table; whether a re-run is in flight is not recorded in the folder (check CAR/scripts/lepton/poll_carey.sh output / Lepton job state before acting).
11. Mixed-assay EM (GBS progeny + WGS parents share dA/dB) is a known formalization issue, not resolved (TECH I-20).

6. Best scripts to generalise (by stage)