This is the working manuscript scaffold. Bracketed [TO CONFIRM] markers denote fields for the authors to finalise.

title: "Reactive-oxygen and calcium signalling shape the Arabidopsis transcriptomic and root-architectural response to spaceflight (APEX-05)" running_title: "ROS and Ca2+ signalling in the APEX-05 spaceflight response" target_journal: "npj Microgravity (Article)" manuscript_status: "DRAFT SCAFFOLD — assembled from released data, storyboard, and reanalysis outputs. Passages marked [TO CONFIRM] require the authors to insert or verify exact values before submission; no quantitative claim here should be treated as final until checked against the source tables." authors: - name: "Richard Barker" affiliation: 1 corresponding: true email: "admin@cosecloud.com" - name: "[TO CONFIRM: co-authors, ORCID iDs, affiliations]" affiliations: - id: 1 name: "[TO CONFIRM: institution, city, country]" keywords: [spaceflight, Arabidopsis thaliana, transcriptomics, root system architecture, reactive oxygen species, RBOHD, calcium, CAX, FAIR data]

Editorial note (remove before submission). This file is a scaffold. Its structure and section flow follow npj Microgravity's Article format. Numerical results are drawn directly from the released tables in this repository and are cited by filename; where a number could not be verified from the released files it is written as [TO CONFIRM]. The machine-learning quality-control results (Fig. 5) were produced by analysis/ml/apex05_ml_anomaly_detection.py in this repository and are reproducible from results/ml/ml_metrics.json.

Abstract

Spaceflight imposes a unique combination of microgravity and space-radiation stress on plants, yet the signalling pathways that translate the orbital environment into altered growth remain incompletely mapped. The Advanced Plant Experiment-05 (APEX-05) flew four Arabidopsis thaliana genotypes — wild-type Col-0, a mutant in reactive-oxygen (ROS) production (rbohD), and two independent alleles of the calcium/H⁺ exchanger CAX2 (cax2-2 and cax2-3) — aboard the International Space Station, alongside matched ground controls, and profiled both the shoot and root transcriptome (RNA-seq) and root system architecture (RSML morphometrics). A QC concordance test between the two CAX2 alleles found cax2-3 strongly discordant with its sibling cax2-2 (DEG set ~10–17× inflated and no more similar to cax2-2 than to unrelated genotypes); cax2-3 was therefore excluded, leaving a primary set of three genotypes (Col-0, cax2-2, rbohD). Wild-type rosette leaf area was reduced under spaceflight [TO CONFIRM: exact means/CI from the leaf-area analysis; storyboard indicates ~4.0 cm² ground control vs ~2.7 cm² flight], and this response was altered in the ROS- and calcium-signalling mutants, implicating both pathways in the spaceflight growth response. Differential-expression analysis identified hundreds of flight-responsive genes in each genotype and tissue (e.g. Col-0 root: 799 DEGs, 268 flight-induced / 531 flight-repressed; see Table 1 and the direction note in §2), enriched for photosynthesis, stress-response, cell-wall and redox pathways (Fig. 9; results/tables/apex05_primary_enrichment_*). Beyond the biology, we release APEX-05 as a FAIR, reproducible data package and demonstrate a machine-learning quality-control workflow that (i) recovers tissue identity from expression with 100% leave-one-out accuracy and detects deliberately mislabelled samples at ~97% recall, and (ii) shows that day-4 primary-root morphology only weakly distinguishes these subtle mutants — the finding that motivated our provenance- based correction of a historical cax23 sample-labelling anomaly. Together the data and tooling provide a reusable resource for the plant space-biology community.

Introduction

[Scaffold — 3–4 paragraphs. Suggested content, to be written/verified by authors:]

The problem. Plants are candidate life-support organisms for long-duration spaceflight, but microgravity perturbs gravitropic, hydrotropic, oxidative and cell-wall processes. Prior spaceflight transcriptomics (e.g. the GeneLab / Open Science Data Repository corpus) established broad reprogramming but the causal signalling nodes are debated. [TO CONFIRM: citations.]
The candidate pathways. Reactive oxygen species produced by the NADPH oxidase RBOHD, and calcium fluxes mediated by CAX-family exchangers, are central to plant environmental sensing and have been proposed as early relays in the spaceflight response. APEX-05 was designed to test their contribution directly by flying rbohD and two independent alleles of the same gene CAX2 — cax2-2 and cax2-3 — against Col-0. Including two alleles of one gene provides a built-in reproducibility control: concordant responses corroborate a genuine CAX2 phenotype, whereas divergence flags an allele- or sample-specific artefact. [TO CONFIRM: precise allele identifiers / T-DNA insertion lines and prior hypotheses.]
This study. We report the paired shoot/root transcriptome and root- architecture response, and — reflecting community priorities around reproducibility — package the analysis to FAIR standards and add an independent machine-learning QC layer. We use that layer to document and resolve a sample-labelling anomaly that arose because cax22 and cax23 were grown in physically adjacent plate wells.

Results

1. Spaceflight reduces wild-type leaf area and the mutants respond differently

Rosette leaf area measured at [TO CONFIRM: 8 days] showed a reduction in wild-type Col-0 under spaceflight relative to ground control (Fig. 1). The ROS-signalling mutant rbohD and the calcium-transport mutants cax22/cax23 displayed [TO CONFIRM: describe the direction and magnitude of the genotype × environment interaction from the leaf-area dataset]. This phenotype frames the transcriptomic analysis that follows: if ROS and Ca²⁺ signalling gate the growth response, the flight-responsive transcriptomes of these mutants should diverge from wild type.

Source: reports/apex05_storyboard.png; leaf-area measurements [TO CONFIRM: add the underlying leaf-area table to data/ and cite it here].

2. Hundreds of genes respond to spaceflight in each genotype and tissue

Flight-versus-ground-control differential expression (edgeR; [TO CONFIRM: thresholds, e.g. FDR < 0.05, |log2FC| > 1]) identified robust, tissue-specific responses (Table 1, Fig. 2). In wild-type Col-0 we detected 799 flight- responsive genes in root (268 induced / 531 repressed by flight) and 604 in shoot (201 induced / 403 repressed). The mutants showed responses of comparable scale but distinct composition: rbohD (root 626: 233 induced / 393 repressed; shoot 456: 286 induced / 170 repressed) and cax2-2 (root 324: 223 induced / 101 repressed; shoot 230: 185 induced / 45 repressed).

Direction convention. Throughout this manuscript, differential-expression directions are flight-referenced: induced = higher in spaceflight (FL), repressed = higher in ground control (GC). Note that the underlying released DEG workbooks are named *_up-regulated_* / *_down-regulated_* on the opposite, ground-control-referenced basis — their "up-regulated" genes are 100% higher in GC (i.e. flight-repressed), verified directly from the labelled baseMeanA_FL / baseMeanB_GC columns. We re-derived every direction from those columns so that the counts and enrichment reported here follow the flight-referenced convention.

Source tables: results/tables/apex5_col_root_edger_*, apex5_col_shoot_edger_*, apex5_rbohd_root_*, apex5_rbohd_shoot_*, apex5_cax22_root_*, apex5_cax22_shoot_*, and the combined apex5_edger_deg_all-samples.csv.

Genotype-resolved over-representation analysis of these flight-DEG sets (g:Profiler, g:SCS-corrected p < 0.05, tissue-specific tested-gene background; Fig. 9, analysis/ml/apex05_primary_genotype_enrichment.py) revealed distinct functional signatures. Wild-type Col-0 mounted a broad stress/stimulus and transmembrane-transport response in both organs (root "response to chemical" p = 4×10⁻²²; flight-induced aquaporin/water-transport genes). rbohD was distinguished by a redox/detoxification signature (glutathione-transferase and oxidoreductase activity, response to toxic substance; shoot glutathione and glucosinolate metabolism), consistent with its role in ROS production. cax2-2 was the most distinctive: its flight-induced genes were dominated by photosynthesis/thylakoid in root (p = 5×10⁻⁵⁶) and cell-wall organisation and phenylpropanoid biosynthesis in shoot — reproducing, from the retained allele alone, the CAX2 concordant-core signature of §6. Full term tables: results/tables/apex05_primary_enrichment_{col0,cax22,rbohd}_{root,shoot}.csv.

3. Root and shoot deploy largely distinct flight-responsive programmes

Comparison of root versus shoot responses (Fig. 3; jVenn/UpSet in results/plots/apex05_col-cax22-cax23_upset.svg and results/tables/apex5_roots-vs-shoots_jvenn_transcripts.csv) showed [TO CONFIRM: the overlap statistics — number of shared vs organ-specific DEGs]. Functional enrichment (Table 2) implicated [TO CONFIRM: top terms from apex5_root-shoot_edger_enrichment_*.tsv, apex5_shoots_core-enrichment.csv], consistent with organ-specialised remodelling of [TO CONFIRM: e.g. cell-wall, photosynthesis, defence] processes.

4. Cell-wall and ROS pathways are enriched in the rbohD response

Mapping the rbohD flight response onto AraCyc/KEGG pathways (results/plots/apex05_rbohd_roots_aracyc.png, data/genesets/apex05_rbohd_aracyc_cellwall.txt) highlighted cell-wall biosynthesis and [TO CONFIRM] pathways, linking altered ROS production to structural remodelling under spaceflight. [TO CONFIRM: quantitative pathway enrichment statistics.]

5. Machine-learning quality control detects mislabelled samples and validates the corrected design

Because cax2-2 and cax2-3 were grown in physically adjacent plate wells, an early QC screen flagged a cax2-3 labelling anomaly (the "CAX23oddity"; full history in docs/PROVENANCE_cax23_mislabelling.md). We built an independent machine-learning check (analysis/ml/apex05_ml_anomaly_detection.py, Fig. 5):

Expression carries an unambiguous tissue signal. A leave-one-out classifier separated Col-0 root from shoot transcriptomes with 100% accuracy (16 samples, 8,068 shared genes). Injecting known label swaps and asking the detector to recover them yielded ~97% recall (top-2 of 200 trials) with near-complete score separation between swapped (mean 0.98) and correctly-labelled (mean 0.09) samples (Fig. 5a,b). This is the modality in which a mislabelled sample is detectable — the same principle that surfaced the original oddity.
Root morphology alone cannot re-derive genotype. A leave-one-well-out classifier on aggregated RSML primary-root features reached only ~44% genotype accuracy (chance 25%; Fig. 5c,d), with the genotypes overlapping in morphometric space. This negative result is informative: because these subtle ROS/Ca²⁺ mutants are not phenotypically separable at day 4, labels cannot be rescued from phenotype, and the correct remedy is provenance-based — we re-keyed every sample to its plate-well ID, removing the ambiguous cax22 well A7 and the anomalous cax23 ground-control replicate to yield the clean 60-sample master design.

All metrics are reproducible from results/ml/ml_metrics.json (seed 42).

6. The two CAX2 alleles are transcriptionally discordant

Because cax2-2 and cax2-3 disrupt the same gene, their spaceflight responses should agree. Instead, a concordance analysis of the released DEG sets (analysis/ml/apex05_cax2_allele_concordance.py, Fig. 6) found cax2-3 with a grossly inflated flight-response — 4,539 root / 2,980 shoot DEGs versus 473 / 174 for cax2-2, i.e. 9.6× / 17.1× larger and exceeding every other genotype including wild type (Fig. 6a). Critically, cax2-2 overlapped its own sibling allele no more than it overlapped the unrelated genotypes (DEG-set Jaccard 0.069 root / 0.045 shoot vs 0.072–0.118 for Col-0 and rbohD; Fig. 6b). The genuine shared core is small but real: of genes DE in both alleles (216 root, 151 shoot), fold-change direction agreed 97%. This pattern — a modest direction-concordant core beneath a large, non-preferential cax2-3-exclusive excess — indicates the cax2-3 libraries carry substantial technical variance rather than a real allelic divergence, consistent with the adjacent-well anomaly. We therefore excluded cax2-3 from the primary analysis and report cax2-2 as the CAX2 representative (retained cax2-3 result tables: archive/excluded_cax2-3/; decision record: docs/cax2_allele_concordance.md). cax2-3 is retained only within the QC analyses (this concordance test and the morphometric confusion of Fig. 5) that justify its exclusion.

To preserve the corroborated part of the CAX2 signal, we additionally define the CAX2 concordant core — genes differentially expressed under spaceflight in both alleles with agreeing direction (Fig. 7, analysis/ml/apex05_cax2_concordant_core.py). This yields 210 root / 146 shoot high-confidence CAX2 flight-responsive genes (of 216 / 151 shared; 97% direction agreement). The core is overwhelmingly flight-induced (≈97% up) and entirely tissue-partitioned (no gene shared between the root and shoot cores), and is released as a supplementary table (results/tables/apex05_cax2_concordant-core_*.csv) as the defensible CAX2 flight signature.

Over-representation analysis of the core (g:Profiler, g:SCS-corrected, against the tissue-specific tested-gene background; Fig. 8, analysis/ml/apex05_cax2_core_enrichment.py) shows the two tissues respond through distinct programmes. The (flight-induced) root core is dominated by photosynthesis / thylakoid terms (GO:CC thylakoid p = 9×10⁻⁵⁸; photosynthesis, photosystem I), indicating ectopic induction of the photosynthetic apparatus in roots under spaceflight. The shoot core is enriched for cell-wall organisation and phenylpropanoid biosynthesis (plant-type cell wall organisation p = 1×10⁻¹⁰; KEGG phenylpropanoid biosynthesis; structural constituent of cell wall) together with peroxidase / hydrogen-peroxide-catabolic terms (lactoperoxidase activity, hydrogen peroxide catabolic process, heme binding) — connecting the CAX2 flight response to cell-wall remodelling and the ROS axis also probed by rbohD. Full term lists: results/tables/apex05_cax2_core_enrichment_{root,shoot}.csv.

Discussion

[Scaffold — to be written by authors. Key points supported by this repository:]

Biology. ROS (RBOHD) and calcium (CAX) signalling both modulate the Arabidopsis spaceflight response, evidenced by genotype-specific leaf-area phenotypes and divergent flight-responsive transcriptomes. [TO CONFIRM: integrate with prior spaceflight ROS/Ca²⁺ literature and OSDR datasets.]
Organ specialisation. Root and shoot mount largely separate programmes, arguing against a single systemic flight signal. [TO CONFIRM.]
Reproducibility as a result. Adjacent-well genotype confusion is a generic risk in multi-genotype spaceflight hardware; keying samples on physical position and validating with an ML QC layer is a transferable practice for the field.
Limitations. Per-sample expression matrices are released here for Col-0 only; the mutant contrasts are provided as summary/DEG tables. [TO CONFIRM: note replicate numbers, power, and any batch structure.]

Methods

[Scaffold — expand/verify. Anchor each subsection to the released files.]

Plant material and spaceflight hardware. Four genotypes (Col-0, rbohD, cax22, cax23); [TO CONFIRM: growth hardware, light, media, flight profile, fixation]. Sample design in data/metadata/apex05_sample_metadata.csv (60 samples; genotype × tissue × flight/ground × replicate, keyed on plate-well ID).
RNA-seq and read processing. HiSeq sequencing; HISAT2 alignment; targets/BAM maps in data/metadata/ (see also archived Targets File revisions in archive/targets_file_versions/). [TO CONFIRM: genome build, aligner version, counting.]
Differential expression. edgeR and DESeq2 (analysis/R/), FL vs GC per genotype × tissue; [TO CONFIRM: thresholds]. Outlier-handling variant in analysis/R/apex5_edger_outlier-removed-main-analysis.R. All directions are reported flight-referenced (induced = higher in FL), re-derived from the labelled baseMeanA_FL / baseMeanB_GC columns; the archived DEG workbooks' up/down file labels are ground-control-referenced and therefore inverted.
Functional enrichment. Manuscript enrichment (Figs 8–9) uses g:Profiler g:GOSt (organism A. thaliana; g:SCS-corrected p < 0.05; custom tissue-specific tested-gene background) via analysis/ml/apex05_cax2_core_enrichment.py and analysis/ml/apex05_primary_genotype_enrichment.py; outputs in results/tables/apex05_*_enrichment_*.csv. Legacy GO/KEGG/AraCyc exports (iDEP/Metascape/pathview) are retained in results/tables/apex5_*enrichment* and data/genesets/.
Root system architecture. RSML exported from Fiji/archidart (analysis/R/apex5_drb_rsml-analysis.R, data/morphometrics/apex05_rsml_day4_morphometrics.csv). [TO CONFIRM: imaging, day, traits.]
Machine-learning QC. analysis/ml/apex05_ml_anomaly_detection.py (scikit-learn; leave-one-out / leave-one-well-out CV; IsolationForest; seed 42). See analysis/ml/README.md.
Sample-label provenance. The cax23 anomaly and its resolution are documented in docs/PROVENANCE_cax23_mislabelling.md.

Data availability

The complete data package — metadata, expression tables, root morphometrics, gene sets, results and figures — is in this repository and archived at Zenodo (DOI: [TO CONFIRM after Zenodo deposit]). Raw sequencing reads are available from NASA OSDR / GeneLab under accession [TO CONFIRM: OSD-###].

Code availability

All analysis code (R differential-expression/pathway pipelines and the Python ML QC pipeline) is in analysis/, released under the MIT License. The ML results are fully reproducible via analysis/ml/apex05_ml_anomaly_detection.py.

Acknowledgements / Author contributions / Competing interests

[TO CONFIRM.]

References

[TO CONFIRM — populate from the Introduction/Discussion citations.]

APEX-05 — Figure legends

Legends for the main-text and machine-learning QC figures. Each entry names the source file(s) in this repository so figures can be regenerated or traced. Values marked [TO CONFIRM] must be verified against the source data before submission.

Main figures

Figure 1 | Spaceflight reduces wild-type leaf area and the ROS/Ca²⁺ mutants respond differently. Violin plots of rosette leaf area (cm²) at [TO CONFIRM: 8 days] for Col-0, rbohD, cax22 (cax2.2) and cax23 (cax2.3) under ground control (GC) and spaceflight (FL). Horizontal bars, group means; dashed line, [TO CONFIRM: grand mean / wild-type GC reference]. Wild-type leaf area is reduced under flight (~4.0 → ~2.7 cm² [TO CONFIRM]); the mutants show an altered genotype × environment response. n = [TO CONFIRM] plants per group. Source: reports/apex05_storyboard.png (regenerate from the underlying leaf-area table once added to data/).

Figure 2 | Flight-responsive differential expression across genotypes and tissues. [TO CONFIRM: chosen visualisation — e.g. bar plot of up/down DEG counts per genotype × tissue, or volcano panels.] DEG counts (FL vs GC; [TO CONFIRM: FDR/|log2FC| cut-offs]): Col-0 root 531↑/268↓, Col-0 shoot 403↑/201↓, rbohD root 393↑/233↓, rbohD shoot 170↑/286↓, cax22 root 100↑/223↓, cax22 shoot 45↑/185↓. Source: results/tables/apex5_*_edger_*regulated-*.xlsx, results/tables/apex5_edger_deg_all-samples.csv.

Figure 3 | Root and shoot deploy largely distinct flight-responsive programmes. UpSet / Venn comparison of up- and down-regulated gene sets between root and shoot (and across genotypes). [TO CONFIRM: describe the largest intersections and the organ-specific fractions.] Source: results/plots/apex05_col-cax22-cax23_upset.svg, results/plots/apex05_roots-vs-shoots_jvenn_transcripts.png, results/tables/apex5_roots-vs-shoots_jvenn_transcripts.csv.

Figure 4 | Cell-wall and ROS pathways in the rbohD flight response. AraCyc/KEGG pathway projection of the rbohD root flight response, highlighting cell-wall biosynthesis and [TO CONFIRM] pathways. [TO CONFIRM: legend detail on colour scale = log2 fold change.] Source: results/plots/apex05_rbohd_roots_aracyc.png, data/genesets/apex05_rbohd_aracyc_cellwall.txt.

Figure 5 | Machine-learning quality control detects mislabelled samples and validates the corrected design. Produced by analysis/ml/apex05_ml_anomaly_detection.py (scikit-learn; seed 42). (a) PCA of Col-0 transcriptomes (log-CPM, 8,068 genes) coloured by tissue; root and shoot separate completely (leave-one-out classifier accuracy 100%). (b) Distribution of the per-sample mislabel score (1 − P[stated label]) over 200 controlled label-swap injections; deliberately swapped samples (mean 0.98) separate cleanly from correctly-labelled samples (mean 0.09), recovered at ~97% recall. (c) Leave-one-well-out genotype confusion matrix from aggregated day-4 primary-root RSML morphometrics (48 well × treatment profiles); overall accuracy ~44% (chance 25%), i.e. genotypes overlap in morphometric space. (d) PCA of the aggregated morphometric space by genotype, showing the same overlap. Panels (c–d) establish that labels cannot be rescued from phenotype, motivating the provenance-based (well-ID) correction described in the text. Source panels: results/ml/figA1_tissue_expression_pca.png, figA2_expression_recovery_scores.png, figB1_genotype_confusion_matrix.png, figB2_morphometric_pca.png; metrics in results/ml/ml_metrics.json.

Figure 6 | The two CAX2 alleles are transcriptionally discordant. Produced by analysis/ml/apex05_cax2_allele_concordance.py. cax2-2 (cax22) and cax2-3 (cax23) are two independent mutant alleles of the same gene, CAX2, and should respond alike to spaceflight. (a) Flight-responsive DEG set size per genotype (root and shoot): cax2-3 is inflated 9.6× (root) / 17.1× (shoot) relative to cax2-2 and exceeds all other genotypes. (b) Pairwise DEG-set Jaccard heatmaps (root, shoot); cax2-2 overlaps cax2-3 no more than it overlaps the unrelated Col-0 and rbohD. Of genes DE in both alleles, fold-change direction agrees 97%, indicating a small genuine shared core beneath a large cax2-3-specific excess. Interpreted as technical/background variance in the cax2-3 libraries rather than a real allelic divergence. Source panels: results/ml/figC1_cax2_deg_set_sizes.png, figC2_cax2_jaccard.png; metrics in results/ml/cax2_concordance_metrics.json; full write-up in docs/cax2_allele_concordance.md.

Figure 7 | CAX2 concordant core — the high-confidence flight signature. Produced by analysis/ml/apex05_cax2_concordant_core.py. Per-gene spaceflight log2 fold-change (FL/GC) in cax2-2 (x) vs cax2-3 (y) for genes differentially expressed in both alleles, root (left) and shoot (right). cax2-2's direction is computed from its labelled baseMeanA_FL/baseMeanB_GC columns; cax2-3's orientation is resolved empirically against cax2-2 (it is flipped — its raw log2FoldChange is log2 GC/FL). Orange = concordant core (agreeing direction: 210 root / 146 shoot of 216 / 151 shared, 97%); grey = discordant. Points lie predominantly in the upper-right (flight-induced in both) and above the identity line (cax2-3 magnitudes inflated, but directionally concordant). The core is released as results/tables/apex05_cax2_concordant-core_{root,shoot,combined}.csv.

Figure 8 | Functional enrichment of the CAX2 concordant core. Produced by analysis/ml/apex05_cax2_core_enrichment.py (g:Profiler g:GOSt, organism A. thaliana, g:SCS-corrected p < 0.05, custom background = the tissue-specific set of genes tested for differential expression). Top terms (bars = −log₁₀ adjusted p, coloured by source: GO:BP/MF/CC, KEGG, Reactome, WikiPathways) for the root core (left; 92 significant terms) and shoot core (right; 32 terms). The root core is dominated by photosynthesis/thylakoid terms; the shoot core by cell-wall organisation, phenylpropanoid biosynthesis and peroxidase/H₂O₂-catabolic terms. Full tables: results/tables/apex05_cax2_core_enrichment_{root,shoot,all}.csv.

Figure 9 | Genotype-resolved functional enrichment of the flight response. Produced by analysis/ml/apex05_primary_genotype_enrichment.py (g:Profiler g:GOSt, A. thaliana, g:SCS-corrected p < 0.05, custom tissue-specific tested-gene background). Top enriched terms (−log₁₀ adjusted p, coloured by source) for the union flight-DEG set of each primary genotype — Col-0, cax2-2, rbohD (columns) — in root (top) and shoot (bottom). Directions used throughout are flight-referenced (see §2 direction note). Col-0 shows a broad stress/transport response; rbohD a redox/detoxification signature; cax2-2 a photosynthesis (root) / cell-wall (shoot) signature that mirrors the CAX2 concordant core (Fig. 8). Full tables: results/tables/apex05_primary_enrichment_*.csv.

Supplementary figures

Figure S1 | CAX23oddity diagnostic that first flagged the labelling anomaly. jVenn/UpSet DEG set-membership across the four root genotypes; cax23's overlap pattern is inconsistent with a clean genotype cluster. Source: archive/diagnostics/apex05_roots_v11_cax23-oddity.png, results/plots/apex05_cax23_shoot_v11_cax23-oddity.png.

Figure S2 | WT-vs-rbohD differential expression (jVenn). Root and leaf/shoot overlaps between wild type and rbohD. Source: results/plots/apex05_root_wt-vs-rbohd_jvenn.png, results/plots/apex05_leaf_wt-vs-rbohd_jvenn.png.

Figure S3 | Per-genotype/pathway summary panels. [TO CONFIRM: curate from results/plots/apex5_col0.png, apex5_cipk20.png, apex5_genotype-plot.png, apex5_upsetr-plot.png, and the DESeq/edgeR summary PDFs in archive/superseded/.]

Figure S4 | Root system architecture example and time-lapse. Representative Col-0 GC/FL root imaging. Source: media/apex05_col0_gc-fl_timelapse_1080p.mov, data/morphometrics/apex05_rsml_source_col-cax22.zip.