arxiv: 2604.02942 · v1 · submitted 2026-04-03 · 💻 cs.LG

Recognition: 1 theorem link

· Lean Theorem

Explainable Machine Learning Reveals 12-Fold Ucp1 Upregulation and Thermogenic Reprogramming in Female Mouse White Adipose Tissue After 37 Days of Microgravity: First AI/ML Analysis of NASA OSD-970

Md. Rashadul Islam

Authors on Pith no claims yet

Pith reviewed 2026-05-13 20:21 UTC · model grok-4.3

classification 💻 cs.LG

keywords microgravityUcp1thermogenesiswhite adipose tissueexplainable AIRandom Forestgene expressionNASA OSD-970

0 comments

The pith

Microgravity triggers 12-fold Ucp1 upregulation in female mouse white fat, activating thermogenesis as a compensatory response.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper re-analyzes gene expression data from 16 female mice after 37 days aboard the ISS using differential expression tests, Random Forest classifiers, and SHAP explanations on 89 genes tied to fat development and heat production. It reports a 12.21-fold rise in Ucp1 transcript levels together with broad activation of the thermogenesis pathway. A sympathetic reader cares because the result points to rapid metabolic rewiring in white fat under weightlessness, with possible carry-over to astronaut physiology and to metabolic disorders studied on Earth.

Core claim

After 37 days of microgravity, gonadal white adipose tissue from female C57BL/6J mice exhibits a 12.21-fold upregulation of Ucp1 (ΔΔCt = -3.61, p = 0.0167) accompanied by a mean 3.24-fold activation of the thermogenesis gene set. Random Forest models trained on the top-20 genes separate flight from ground-control samples with AUC 0.922 under leave-one-out cross-validation, and SHAP values consistently place Ucp1 at the top of the feature ranking while Angpt2, Irs2, Jun, and Klf transcription factors appear as secondary consensus drivers. Principal component analysis shows flight and control groups cleanly separated along PC1, which accounts for 69.1 percent of variance.

What carries the argument

Random Forest classification with SHAP explanations applied to RT-qPCR measurements of 89 adipogenesis and thermogenesis genes, isolating Ucp1 upregulation as the primary marker of thermogenic reprogramming.

If this is right

Thermogenic reprogramming in white adipose tissue may act as a rapid compensatory mechanism to microgravity-induced metabolic stress.
The molecular signature could inform countermeasures for female astronauts on long-duration missions.
Similar reprogramming patterns may be relevant to ground-based research on obesity and type-2 diabetes where white fat browning is a therapeutic target.
PCA separation indicates that a compact gene panel can reliably distinguish microgravity-exposed tissue from controls.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

If the Ucp1-driven shift proves general, it may appear in other fat depots or in male mice under comparable exposure.
Targeted interventions such as timed exercise or temperature control during flight could be tested to blunt or enhance the reprogramming.
The same explainable-ML workflow could be applied to newly released NASA datasets on other tissues to map whole-body metabolic responses.
Confirmation in human cell or organoid models would strengthen translational value for both space and metabolic medicine.

Load-bearing premise

The observed gene-expression shifts are caused by microgravity exposure itself rather than launch stress, radiation, or other flight variables.

What would settle it

A follow-up flight experiment that adds ground controls subjected to identical launch vibration and radiation profiles would show whether the 12-fold Ucp1 increase disappears when those confounds are matched.

Figures

Figures reproduced from arXiv: 2604.02942 by Md. Rashadul Islam.

**Figure 2.** Figure 2: Principal Component Analysis of 89-gene Ct profiles. PC1 (69.1%) and PC2 (11.7%) together explain 80.8% of variance. Flight and [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: Volcano plot of differential gene expression in gonadal WAT [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

**Figure 4.** Figure 4: UCP1 deep-dive analysis. (A) Individual Ct values by group: flight (mean Ct [PITH_FULL_IMAGE:figures/full_fig_p006_4.png] view at source ↗

**Figure 5.** Figure 5: ROC curves for all classifiers under LOO-CV (top-20 gene feature set). The Random Forest and Logistic Regression models achieved [PITH_FULL_IMAGE:figures/full_fig_p007_5.png] view at source ↗

**Figure 6.** Figure 6: SHAP summary bar plots for (left) Random Forest and (right) XGBoost. Bar height represents mean [PITH_FULL_IMAGE:figures/full_fig_p007_6.png] view at source ↗

**Figure 7.** Figure 7: SHAP beeswarm plots for Random Forest (top) and XGBoost (bottom). Each point represents one sample; color encodes gene [PITH_FULL_IMAGE:figures/full_fig_p008_7.png] view at source ↗

**Figure 8.** Figure 8: Consensus feature importance across all models (top 20 genes). Left: bar chart of consensus scores. Right: individual model importance [PITH_FULL_IMAGE:figures/full_fig_p008_8.png] view at source ↗

**Figure 9.** Figure 9: Correlation matrix and network for the top-25 consensus genes. The Angpt2–Jun–Irs2–Klf2 module (upper-left cluster) shows strong [PITH_FULL_IMAGE:figures/full_fig_p009_9.png] view at source ↗

read the original abstract

Microgravity induces profound metabolic adaptations in mammalian physiology, yet the molecular mechanisms governing thermogenesis in female white adipose tissue (WAT) remain poorly characterized. This paper presents the first machine learning (ML) analysis of NASA Open Science Data Repository (OSDR) dataset OSD-970, derived from the Rodent Research-1 (RR-1) mission. Using RT-qPCR data from 89 adipogenesis and thermogenesis pathway genes in gonadal WAT of 16 female C57BL/6J mice (8 flight, 8 ground control) following 37 days aboard the International Space Station (ISS), we applied differential expression analysis, multiple ML classifiers with Leave-One-Out Cross-Validation (LOO-CV), and Explainable AI via SHapley Additive exPlanations (SHAP). The most striking finding is a dramatic 12.21-fold upregulation of Ucp1 (Delta-Delta-Ct = -3.61, p = 0.0167) in microgravity-exposed WAT, accompanied by significant activation of the thermogenesis pathway (mean pathway fold-change = 3.24). The best-performing model (Random Forest with top-20 features) achieved AUC = 0.922, Accuracy = 0.812, and F1 = 0.824 via LOO-CV. SHAP analysis consistently ranked Ucp1 among the top predictive features, while Angpt2, Irs2, Jun, and Klf-family transcription factors emerged as dominant consensus classifiers. Principal component analysis (PCA) revealed clear separation between flight and ground samples, with PC1 explaining 69.1% of variance. These results suggest rapid thermogenic reprogramming in female WAT as a compensatory response to microgravity. This study demonstrates the power of explainable AI for re-analysis of newly released NASA space biology datasets, with direct implications for female astronaut health on long-duration missions and for Earth-based obesity and metabolic disease research.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

This is the first ML re-analysis of the OSD-970 NASA dataset and it reports a large 12-fold Ucp1 rise in female mouse WAT after 37 days in space, but the causal link to microgravity alone is not yet locked down.

read the letter

The main thing to know is that this paper takes the public OSD-970 RT-qPCR data from 16 female mice and runs it through standard classifiers plus SHAP, turning up a 12.21-fold Ucp1 increase and thermogenesis pathway activation that had not been quantified this way before. That is the concrete new piece. They also get a Random Forest AUC of 0.922 under leave-one-out cross-validation and show clean PCA separation, which is decent for a small dataset of 89 genes. Re-using open NASA space biology data with explainable ML is a reasonable move and the SHAP rankings (Ucp1, Angpt2, Klf factors) give a clear picture of what the model is using. That part is useful for anyone who wants to see how these tools land on rodent flight samples. The soft spots are straightforward. The sample is only 8 flight versus 8 ground controls, top-20 feature selection happens after differential expression, and the abstract gives limited preprocessing details. More critically, the ground controls do not obviously rule out launch g-forces, radiation, or feeding shifts as drivers of the gene changes. Without extra covariates or parallel controls like hindlimb unloading, the jump to “microgravity-induced thermogenic reprogramming” rests on an assumption that still needs testing. The math and data handling look standard rather than inventive, and the citation pattern is thin because the work is mostly re-analysis. This is for people who follow space physiology or small-omics ML applications. A reader who wants a quick example of SHAP on NASA rodent data will get value, but anyone needing firm causal evidence should wait for tighter controls. I would send it to peer review so referees can check the confounder handling and whether the fold-change survives stricter validation.

Referee Report

3 major / 2 minor

Summary. The manuscript presents the first machine learning analysis of NASA OSD-970 dataset from the RR-1 mission, using RT-qPCR data on 89 genes from gonadal white adipose tissue of 16 female mice (8 microgravity-exposed, 8 ground controls). It reports a 12.21-fold upregulation of Ucp1 (p=0.0167) and activation of the thermogenesis pathway (mean fold-change 3.24), with Random Forest achieving AUC 0.922 via LOO-CV on top-20 features selected after differential expression, and SHAP identifying Ucp1 and other genes as key predictors. The authors interpret this as evidence of microgravity-induced thermogenic reprogramming in female WAT.

Significance. If the central findings hold after addressing methodological concerns, the work would be significant for space biology by identifying rapid thermogenic adaptations in female WAT during spaceflight, with potential implications for astronaut health on long-duration missions and for metabolic disease research on Earth. The application of SHAP explainability to a public NASA dataset is a clear strength, offering a reproducible template for re-analysis of space omics data.

major comments (3)

[Methods (ML pipeline)] Methods (ML pipeline and feature selection): Top-20 feature selection after differential expression on the full set of 89 genes, prior to LOO-CV, introduces post-selection bias that likely inflates the reported AUC=0.922, Accuracy=0.812, and F1=0.824. Nested cross-validation or reporting performance on the full gene set is needed to support the classifier claims.
[Results (Ucp1 and pathway analysis)] Results (Ucp1 upregulation and pathway activation): The attribution of the 12.21-fold Ucp1 change (Delta-Delta-Ct = -3.61) and mean thermogenesis pathway fold-change of 3.24 specifically to microgravity is not supported, as the 8-vs-8 flight vs. ground comparison does not include covariates or parallel controls for launch g-forces, radiation, or feeding changes in the RR-1 mission.
[Statistical analysis] Statistical analysis (sample size and multiple testing): With n=16 total and 89 genes tested, the p=0.0167 for Ucp1 requires explicit multiple-testing correction; the small n also undermines reliability of LOO-CV estimates and the PCA separation (PC1 explaining 69.1% variance) without reported permutation tests or power analysis.

minor comments (2)

[Abstract] Abstract: The claim of being the 'first AI/ML analysis' of OSD-970 should be qualified by checking and citing any prior non-ML publications on the same dataset.
[Figures] Figures: Ensure all panels report exact sample sizes (n=8 per group), error bars (e.g., SEM or SD), and the specific statistical tests used for the fold-changes and pathway means.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which have helped us identify areas for improvement in methodological rigor and interpretation. We provide point-by-point responses below and have revised the manuscript accordingly.

read point-by-point responses

Referee: Methods (ML pipeline and feature selection): Top-20 feature selection after differential expression on the full set of 89 genes, prior to LOO-CV, introduces post-selection bias that likely inflates the reported AUC=0.922, Accuracy=0.812, and F1=0.824. Nested cross-validation or reporting performance on the full gene set is needed to support the classifier claims.

Authors: We acknowledge this valid concern regarding post-selection bias. In the revised manuscript, we will implement nested cross-validation, performing feature selection (top-20 after differential expression) strictly within each outer training fold. We will also report classifier performance metrics using the full set of 89 genes without prior selection as a baseline. These changes will provide more conservative and unbiased estimates while preserving the utility of the explainable ML approach. revision: yes
Referee: Results (Ucp1 upregulation and pathway activation): The attribution of the 12.21-fold Ucp1 change (Delta-Delta-Ct = -3.61) and mean thermogenesis pathway fold-change of 3.24 specifically to microgravity is not supported, as the 8-vs-8 flight vs. ground comparison does not include covariates or parallel controls for launch g-forces, radiation, or feeding changes in the RR-1 mission.

Authors: We agree that the flight-versus-ground design cannot fully isolate microgravity from other mission factors. In the revision, we will add an explicit limitations paragraph rephrasing the findings as changes observed in the spaceflight-exposed group relative to ground controls, rather than claiming exclusive causation by microgravity. We will retain the biological interpretation as a hypothesis-generating observation consistent with the available NASA OSD-970 data while noting the need for future studies with additional controls. revision: partial
Referee: Statistical analysis (sample size and multiple testing): With n=16 total and 89 genes tested, the p=0.0167 for Ucp1 requires explicit multiple-testing correction; the small n also undermines reliability of LOO-CV estimates and the PCA separation (PC1 explaining 69.1% variance) without reported permutation tests or power analysis.

Authors: We accept the need for multiple-testing correction. The revised manuscript will apply the Benjamini-Hochberg FDR procedure across the 89 genes and report both raw and adjusted p-values. We will also add a post-hoc power analysis for the differential expression results and permutation testing (1,000 permutations) to evaluate the statistical significance of the observed PCA separation. The LOO-CV results will be supplemented by the nested CV framework described in the first response to improve robustness given the small sample size. revision: yes

Circularity Check

0 steps flagged

No circularity: standard empirical analysis of public NASA dataset

full rationale

The paper applies off-the-shelf differential expression (Delta-Delta-Ct), LOO-CV Random Forest, PCA, and SHAP to the public OSD-970 qPCR dataset. Fold-changes and model metrics are computed directly from the raw counts without any parameter being fitted to a subset and then re-predicted, without self-definitional equations, and without load-bearing self-citations that close the derivation. The reported 12.21-fold Ucp1 change and AUC=0.922 are independent outputs of the data pipeline, not reductions to the paper's own inputs.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 0 invented entities

The paper relies on standard transcriptomic assumptions and public data without introducing new physical entities or many ad-hoc parameters beyond routine feature selection.

free parameters (1)

top-20 feature cutoff
Post-analysis selection of top 20 genes for final model input, chosen after initial differential expression ranking.

axioms (2)

domain assumption Gene expression measurements are independent across biological replicates
Standard assumption underlying differential expression and cross-validation on the 16-mouse cohort.
domain assumption Microgravity is the dominant experimental variable driving observed gene changes
Assumes flight vs ground contrast isolates microgravity effects without major confounds from launch, radiation, or housing.

pith-pipeline@v0.9.0 · 5687 in / 1502 out tokens · 50343 ms · 2026-05-13T20:21:23.586186+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
The most striking finding is a dramatic 12.21-fold upregulation of Ucp1 (ΔΔCt = −3.61, p = 0.0167) … Random Forest with top-20 features achieved AUC = 0.922 … SHAP analysis consistently ranked Ucp1 among the top predictive features

Reference graph

Works this paper leans on

13 extracted references · 13 canonical work pages

[1]

Rodent research on the International Space Station: The first decade,

A. E. Roncaet al., “Rodent research on the International Space Station: The first decade,”Life Sciences in Space Research, vol. 32, pp. 1–15, 2022

work page 2022
[2]

The NASA rodent research program: Current status and future directions,

J. S. Alwoodet al., “The NASA rodent research program: Current status and future directions,”npj Microgravity, vol. 9, no. 1, p. 12, 2023

work page 2023
[3]

Evidence for increased thermogenesis in female C57BL/6J mice housed aboard the International Space Station,

C. P. Wonget al., “Evidence for increased thermogenesis in female C57BL/6J mice housed aboard the International Space Station,”npj Microgravity, vol. 7, no. 1, p. 13, 2021, pMC8213760

work page 2021
[4]

Simulated microgravity led to increased brown adipose tissue activity in rats,

Y. Chenet al., “Simulated microgravity led to increased brown adipose tissue activity in rats,”Acta Astronautica, vol. 160, pp. 538–551, 2019

work page 2019
[5]

Microgravity and thermogenesis: metabolic implications for long-duration spaceflight,

R. Albustanjiet al., “Microgravity and thermogenesis: metabolic implications for long-duration spaceflight,”Frontiers in Physiology, vol. 10, p. 1095, 2019

work page 2019
[6]

Biospecimen retrieval from NASA’s rodent research-1: Tissue preservation and gene expression in space-flown mice,

S. Y. Choiet al., “Biospecimen retrieval from NASA’s rodent research-1: Tissue preservation and gene expression in space-flown mice,”Gravitational and Space Research, vol. 4, no. 1, pp. 1–12, 2016. [Online]. Available: https: //ntrs.nasa.gov/citations/20160012754

work page arXiv 2016
[7]

OSD-970: Evidence for increased thermogenesis in female C57BL/6J mice housed aboard the ISS — white adipose tissue (WAT) data (GLDS- 790),

NASA Open Science Data Repository, “OSD-970: Evidence for increased thermogenesis in female C57BL/6J mice housed aboard the ISS — white adipose tissue (WAT) data (GLDS- 790),” https://doi.org/10.26030/35bt-r894, 2026

work page doi:10.26030/35bt-r894 2026
[8]

NASA open science data repository (OSDR): Enabling open science for space biology,

R. T. Scottet al., “NASA open science data repository (OSDR): Enabling open science for space biology,”Database, vol. 2024, p. baae012, 2024

work page 2024
[9]

Nrf2 contributes to the weight gain of mice during space travel,

T. Suzukiet al., “Nrf2 contributes to the weight gain of mice during space travel,”Communications Biology, vol. 3, no. 1, p. 496, 2020

work page 2020
[10]

Transcriptomics analysis of rodent spaceflight experiments from the NASA GeneLab repository,

H. Copeet al., “Transcriptomics analysis of rodent spaceflight experiments from the NASA GeneLab repository,”npj Micro- gravity, vol. 10, no. 1, p. 11, 2024

work page 2024
[11]

Explainable AI for space omics: Symbolic re- gression and SHAP analysis of rodent muscle transcriptomics,

Y. Liet al., “Explainable AI for space omics: Symbolic re- gression and SHAP analysis of rodent muscle transcriptomics,” Scientific Reports, vol. 13, no. 1, p. 14567, 2023

work page 2023
[12]

Causal inference machine learning on rodent liver omics from the ISS,

J. Casalettoetal., “Causal inference machine learning on rodent liver omics from the ISS,”npj Microgravity, vol. 11, no. 1, p. 8, 2025

work page 2025
[13]

Harmonized machine learning classifica- tion of spaceflight versus ground control transcriptomics across NASA GeneLab datasets,

A. Ilangovanet al., “Harmonized machine learning classifica- tion of spaceflight versus ground control transcriptomics across NASA GeneLab datasets,”Frontiers in Physiology, vol. 15, p. 1345678, 2024

work page 2024