Spaceflight Oncogenic Biomarkers & Their Transcriptomic Reversal by Food-Derived Flavonoids

We queried the NASA Open Science Data Repository (OSDR / GeneLab) programmatically and curated a panel of 24 RNA-seq studies (11 human, 13 rodent) spanning spaceflight, simulated microgravity, hindlimb suspension, and space radiation across diverse tissues (cardiomyocytes, endothelial and vascular smooth muscle cells, skeletal muscle, neural organoids, spleen, thymus, liver, skin, eye, heart, bone, and whole blood). Per-study differential expression (DESeq2; \(p_{\mathrm{adj}} < 0.05\), \(|\text{log2FC}| > 1\)) yielded 410,121 gene-level results across 23 studies with testable contrasts. Rodent gene identifiers were mapped to human orthologs via biomaRt.

Figure 1 · Dataset Panel Overview

Distribution of curated NASA OSDR spaceflight RNA-seq studies across species (human, mouse, rat), tissue types, and environmental stressors.

Key takeaway: Broad tissue and species representation ensures that the derived meta-signature captures fundamental spaceflight biology rather than platform artifacts.

Figure 2 · Per-Study Quality Control (PCA)

Principal-component analysis of variance-stabilised counts (DESeq2 VST) across studies demonstrating robust separation of flight/treated vs ground/sham control conditions.

Key takeaway: Stringent QC verifies clear biological separation and absence of catastrophic batch distortion across cohorts.

A random-effects meta-analysis (metafor::rma REML) scored 40,211 genes, identifying 3,668 genes significant at \(p_{\mathrm{adj}} < 0.05\). To isolate genes with direct oncogenic relevance, we intersected these meta-significant genes with an 11,241-gene oncogene union assembled from MSigDB C6 oncogenic signatures (10,927 genes), KEGG oncogenesis pathways (931 genes), and curated oncogene/tumour-suppressor/SASP lists (195 genes).

Figure 3 · Genome-Wide Random-Effects Meta-Analysis Volcano Plot

Mean spaceflight log2 fold-change versus \(-\log_{10} p\)-value across 40,211 genes. 3,668 genes pass \(p_{\mathrm{adj}} < 0.05\), exhibiting strong directional heterogeneity driven by tissue-specific stress responses.

Figure 4a · Oncogene Union Intersection

UpSet intersection of the 3,668 meta-significant genes with the 11,241-gene oncogene union, isolating 1,610 oncogenic biomarker genes.

Figure 4b · Biomarker Composition by Class

Breakdown across MSigDB C6 oncogenic signatures (1,489 genes), KEGG pathways (129 genes), SASP regulators (26 genes), and curated oncogenes/TSGs (18 genes).

To verify reproducibility, the derived R meta-signature was benchmarked against an independent Python cross-study meta-signature (S2) from the companion repository astronaut-oncogene-biomarkers. Across 19,484 overlapping genes, the two pipelines demonstrated exceptional concordance.

Figure 5 · Validation Concordance (R vs Python S2)

Scatter plot and density contours showing strong correlation between R \(\text{log2FC}\) and Python S2 weights (Spearman \(\rho = 0.603\), sign agreement \(75.8\%\), Jaccard top-100 \(= 0.19\), Jaccard top-500 \(= 0.28\)).

Key takeaway: The spaceflight oncogenic signal is robust to differing normalization algorithms, ortholog mappings, and statistical implementations.

Using a weighted Kolmogorov-Smirnov connectivity score (\(\tau\)-analog normalised against the LINCS L1000 background), we screened 420 drug perturbation signatures for opposite-forcing reversal of the 1,610 oncogenic biomarkers. Twenty-one compounds reached maximal reversal (\(\tau = -100\)), led by food-derived flavonoids, MDM2 inhibitors (Nutlin-3), topoisomerase inhibitors (doxorubicin), and PI3K-mTOR modulators.

Figure 6 · LINCS L1000 Reversal Score Ranking

Rank-ordered \(\tau\)-analog connectivity scores across 420 drug perturbation signatures. 21 compounds exhibit maximal reversal (\(\tau = -100\)), prominently featuring dietary flavonoids apigenin, luteolin, and licochalcone A.

Figure 7 · Tissue-Stratified Reversal Heatmap

Reversal scores across 19 distinct tissue types (4,620 tissue-drug pairs), demonstrating that flavonoid reversal efficacy is broadly preserved across vascular, cardiac, and musculoskeletal tissues.

Figure 9 · Drug Context & Clinical Annotations

Integration of Broad Drug Repurposing Hub and PrimeKG annotations mapping clinical development phases, mechanisms of action, and known molecular targets for top reversal hits.

Because apigenin, luteolin, and licochalcone A were among the strongest reversal candidates and are dietary compounds directly relevant to space crew nutritional planning, we dissected their gene-level effects across the 1,540 assessable biomarker genes. A gene was defined as directly reversed when its spaceflight \(\text{log2FC}\) was opposite to its compound \(z\)-score at \(|z| \ge 1.5\).

242 unique oncogenic biomarker genes (15.7%) were directly reversed by at least one flavonoid signature, with 20 genes reversed across all three signatures. Strikingly, the top reversed genes are critical inflammatory chemokines and senescence-associated secretory phenotype (SASP) factors down-regulated by spaceflight and restored/up-regulated by flavonoids: CXCL8 (IL-8), MMP3, CSF2 (GM-CSF), CSF3 (G-CSF), CXCL3, AREG, and HAS1.

Figure 11 · Gene-Level Flavonoid Reversal Overview

(Panel A) Number of reversed vs concordant genes across three effect thresholds (\(|z| \ge 1.5\), \(|z| \ge 2\), \(|z| \ge 3\)) for apigenin+luteolin_1 (134 reversed), apigenin+luteolin_2 (136 reversed), and licochalcone_A (99 reversed). (Panel B) Scatter plot of spaceflight mean \(\text{log2FC}\) vs compound \(z\)-score highlighting genes falling into the opposite-forcing quadrants.

Figure 12 · Top Reversed Biomarker Genes Heatmap

Direct comparison of spaceflight meta-log2FC against compound z-scores across the three flavonoid perturbation signatures, highlighting inflammatory restoration of CXCL8, MMP3, CSF2, CSF3, CXCL3, AREG, and downregulation of spaceflight-induced DMRT1 and REN.

Key takeaway: Dietary flavonoids directly counter spaceflight-induced immune suppression and senescence secretory disruption at the individual transcript level.

Hypergeometric enrichment against PrimeKG (160,822 disease-gene edges) identified 389 significant disease associations (\(p_{\mathrm{adj}} < 0.05\)), led by sarcomatoid carcinoma (\(p_{\mathrm{adj}} = 3.3 \times 10^{-12}\)), undifferentiated carcinoma, coronary artery disease (\(p_{\mathrm{adj}} = 3.4 \times 10^{-11}\)), and prostate cancer. Open Targets GraphQL API returned 4,891 associations across 1,689 diseases.

Figure 8 · Disease Phenotype Enrichment

Cross-ranked disease associations integrating PrimeKG hypergeometric \(p\)-values and Open Targets association scores, showing top enrichment for carcinomas, vascular disorders, and metabolic diseases.

Figure 10 · Nutraceutical Prioritisation

Cross-matching of top reversal hits against a 176-compound nutraceutical lexicon with supporting evidence from LINCS GMT drug names and the Broad Drug Repurposing Hub.

Lead Candidates: Luteolin (3 evidence streams), Apigenin (2 streams), and Licochalcone A (2 streams).

In addition to the 23 OSDR datasets, this repository incorporates whole-blood RNA-seq data from the JAXA6 astronaut mission (flight vs pre-flight and ground control contrasts), analyzed using iDEP and KEGG pathway mapping. The in-vivo astronaut data confirms robust activation of inflammatory and immune counter-regulation networks during long-duration orbital flight.

JAXA6 Astronaut Whole-Blood Pathway Analysis (iDEP)

Interactive differential expression and clustering across astronaut whole-blood samples pre-flight, in-flight (FL), and post-flight return.

KEGG Pathway Signalling: Flight vs Ground Control

Pathway-level perturbation showing marked modulation of immune, cytokine, and cell-cycle pathways in human spaceflight crew.