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

cs.LGApr 3, 2026

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

Authors: Md. Rashadul Islam

Organizations: Department of Computer Science and Engineering, Daffodil International University, Dhaka, Bangladesh

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.

Figures & tables

Study	Year	Dataset	Method	Key Finding	vs. This Work
Wong et al.	2021	RR-1 BAT+WAT	t-test & fold-change	Ucp1 $\uparrow$ 1.5 $\times$ in BAT; WAT browning	This work confirms & quantifies ( $\sim$ 12 $\times$ via ML/SHAP)
Chen et al.	2019	Rat tail-suspension	qPCR only	BAT activity $\uparrow$ , UCP1 $\uparrow$	Simulated only; this uses real ISS data + ML classification
Li et al.	2023	RR-1/RR-9 muscle	Symbolic regression + SHAP	Synergistic gene pairs	Muscle focus; this is first WAT thermogenesis + LOO-CV
Casaletto et al.	2025	Rodent liver omics	Causal inference ensemble	Gene–phenotype causality	Liver only; this adds female WAT + multimodal ML
Ilangovan et al.	2024	Multi-liver OSDR	Harmonization + ML classifiers	Space vs. ground classification	No SHAP, no thermogenesis focus; this adds both
Present Work	2026	OSD-970 WAT (new)	RF, XGB, NN + SHAP + LOO-CV	Ucp1 $\sim$ 12 $\times\uparrow$ , thermogenesis dominant	First AI/ML on OSD-970; addresses all gaps

TABLE I: Comparison of the Present Study with Related Work

Panel of Figure 2 — Fig. 1: Analytical pipeline for NASA OSD-970 ML analysis. Raw Ct values from 16 female mice (8 flight, 8 ground) are preprocessed, then subjected to differential expression analysis, multi-classifier LOO-CV, SHAP explainability, and consensus feature ranking with pathway annotation.

Fig. 2: Principal Component Analysis of 89-gene Ct profiles. PC1 (69.1%) and PC2 (11.7%) together explain 80.8% of variance. Flight and ground control samples form distinct clusters along PC1, confirming a strong microgravity-driven transcriptional signal.

Gene	FC	log 2 FC	$p$ -value	Regulation
Ucp1	12.21	3.61	0.0167*	UP
Shh	4.22	2.08	0.1086	n.s.
Wnt3a	1.80	0.85	0.1979	n.s.
Dio2	1.79	0.84	0.3128	n.s.
Angpt2	0.25	$-$ 2.00	0.0001**	DOWN

TABLE II: Top Differentially Expressed Genes (OSD-970, Flight vs. Ground Control)

Fig. 3: Volcano plot of differential gene expression in gonadal WAT (Flight vs. Ground Control). The $x$ -axis shows $\log_{2}$ fold-change ( $\Delta\Delta$ Ct) and the $y$ -axis shows $-\log_{10}(p\text{-value})$ . Ucp1 (red) is the sole significantly upregulated gene (FC $=12.21$ , $p=0.0167$ ). Downregulated significant genes are shown in blue. The dashed lines indicate thresholds $|$ log 2 FC $|=0.585$ and $p=0.05$ .

Fig. 4: UCP1 deep-dive analysis. (A) Individual Ct values by group: flight (mean Ct $=28.69$ ) vs. ground (mean Ct $=32.30$ ); lower Ct indicates higher expression. (B) Box plot with individual data points. (C) Fold-change visualization ( $12.21\times$ upregulation). (D) Distribution density plot showing non-overlapping expression distributions.

Model	Features	Accuracy	F1	AUC	MCC
Random Forest	All 89	0.750	0.750	0.859	0.500
Random Forest	Top 20	0.812	0.824	0.922	0.630
XGBoost	All 89	0.938	0.933	0.875	0.882
XGBoost	Top 20	0.938	0.933	0.875	0.882
Gradient Boost.	All 89	0.938	0.941	0.875	0.882
Gradient Boost.	Top 20	0.875	0.875	0.875	0.750

TABLE III: LOO-CV Classification Results: Flight vs. Ground Control

Fig. 5: ROC curves for all classifiers under LOO-CV (top-20 gene feature set). The Random Forest and Logistic Regression models achieved the highest AUC of 0.922. All models substantially outperformed random classification (dashed diagonal).

Fig. 6: SHAP summary bar plots for (left) Random Forest and (right) XGBoost. Bar height represents mean $|$ SHAP value $|$ across all samples. Ucp1 , Angpt2 , and Irs2 dominate feature importance in both models.

Fig. 7: SHAP beeswarm plots for Random Forest (top) and XGBoost (bottom). Each point represents one sample; color encodes gene expression level (red = high Ct / low expression; blue = low Ct / high expression). Positive SHAP values drive “Flight” predictions; negative values drive “Ground” predictions.

Rank	Gene	Score	FC	Direction
1	Angpt2	1.000	0.25	DOWN
2	Irs2	0.438	0.27	DOWN
3	Jun	0.324	0.23	DOWN
4	Klf2	0.305	0.27	DOWN
5	Ucp1	0.267	12.21	UP
6	Cebpa	0.260	0.38	DOWN

TABLE IV: Top 10 Consensus Feature Importance Scores

Fig. 8: Consensus feature importance across all models (top 20 genes). Left: bar chart of consensus scores. Right: individual model importance scores side-by-side. Angpt2 achieves the highest consensus score (1.000), while Ucp1 ranks 5th with a distinctively high fold-change.

Pathway	Gene Count	Mean FC	Max FC	Min $p$	Mean $\log_{2}$ FC
Thermogenesis	5	3.24	12.21	0.017	0.82
Signaling	12	1.18	4.22	0.011	0.04
Transcription Regulation	18	0.41	0.49	0.001	$-$ 1.44
Adipogenesis	20	0.52	0.91	0.003	$-$ 0.86
Metabolism	15	0.61	1.79	0.018	$-$ 0.64
Other	19	0.70	1.45	0.054	$-$ 0.43

TABLE V: Pathway-Level Differential Expression Summary

Fig. 9: Correlation matrix and network for the top-25 consensus genes. The Angpt2–Jun–Irs2–Klf2 module (upper-left cluster) shows strong mutual correlations ( $r>0.78$ ), consistent with a co-regulated transcriptional network. Ucp1 shows moderate positive correlation with other thermogenesis genes ( Dio2 , Ppargc1a ).

Genomics

Genomics uses computational and experimental methods to study genomes, genetic variation, gene regulation, and their links to traits and disease.

See all 125 papers

Gravity-Awareness: Deep Learning Models and LLM Simulation of Human Awareness in Altered Gravity
Oct 29, 2025Bakytzhan Alibekov, Alina Gutoreva, Elisa Raffaella-FerreComputational Cognitive Modeling Human Behavior Simulation
School of IT & Engineering, Kazakh-British Technical University, 050000, Tole bi 59, Almaty, Kazakhstan. · School of Psychological Sciences, Birkbeck University of London, Malet St, London WC1E 7HX, United Kingdom.
A Unified Three-Stage Machine Learning Framework for Diabetes Detection, Subtype Discrimination, and Cognitive-Metabolic Hypothesis Testing
May 13, 2026Vishal Pandey, Ruzina Haque Laskar, Rishav TewariMedical Diagnosis Clinical Prediction
Independent Researcher · London, UK · Research Scientist - B +4
Faster Thermal Profiling of a Lunar Rover with Machine Learning Adapted Finite Difference Model
May 26, 2026Samuel Weber, Zaki Hasnain, Souma ChowdhuryPDE Solving PDE Surrogate Modeling
University at Buffalo · Department of Mechanical and Aerospace Engineering · Zij Sciences +1

Oct 29, 2025J/K move · Enter open · S save

Gravity-Awareness: Deep Learning Models and LLM Simulation of Human Awareness in Altered Gravity

Bakytzhan Alibekov, Alina Gutoreva, Elisa Raffaella-Ferre

School of IT & Engineering, Kazakh-British Technical University, 050000, Tole bi 59, Almaty, Kazakhstan. · School of Psychological Sciences, Birkbeck University of London, Malet St, London WC1E 7HX, United Kingdom.

Computational Cognitive Modeling Human Behavior Simulation

Open paper AskPDF arXiv

Earth s gravity fundamentally shapes human behaviour. The brain encodes this force as an internal model of gravity, enabling the prediction and interpretation of gravitational effects during perception and action. Understanding how this model adapts to altered gravity is critical for predicting human performance in spaceflight. We present a computational framework for modelling neurophysiological adaptation across diverse gravitational environments. The framework has two components trained on open-access data from altered-gravity studies, particularly parabolic flights. The first component (CorticalG) employs a lightweight multilayer perceptron neural network to predict gravity-dependent changes in EEG frequency bands, estimating cortical state under different gravitational loads. The second component (PhysioG) uses independent Gaussian process models to capture broader physiological responses, including heart rate variability, electrodermal activity, and motor control. To complement the quantitative modelling, we simulated subjective experience across gravitational environments using the Large Language Model (LLM) Claude 3.5 Sonnet. Physiological outputs prompted the model to generate narratives describing alertness, bodily awareness, and cognitive state across zero gravity, partial gravity of the Moon and Mars, and hypergravity. This framework provides a novel approach for investigating human adaptation to spaceflight. It offers a predictive tool to assess performance and resilience, supporting the design of future space exploration missions.

Abstract

Genomics

Explore similar work

Gravity-Awareness: Deep Learning Models and LLM Simulation of Human Awareness in Altered Gravity

A Unified Three-Stage Machine Learning Framework for Diabetes Detection, Subtype Discrimination, and Cognitive-Metabolic Hypothesis Testing

Faster Thermal Profiling of a Lunar Rover with Machine Learning Adapted Finite Difference Model