Three computation approaches were used for the analysis of these five datasets. The comparison of:

1) Differentially expressed genes showed conserved gene expression across multiple flights.

2) Weighted Gene Co-expression network analysis identified groups of interacting genes.

3) Sequential analysis using ImpulseDE2 was adapted to work with increasing gravity intensities.

We were successful in finding conserved responses to micro, lunar, and Martian gravity. This success was seen in the results from ImpulseDE2. This analysis tool is usually used to find changes in gene expression of sequential time points. In this meta-analysis, it was used to find expression trends in sequential gravity intensities. ImpulseDE2 found three transcriptional groupings on the gravity gradient. These groupings of genes are expressed at low levels of gravity, <0.18g, Earth like gravity, >0.65g, and the intermediate gravity levels between those points, 0.18 < intermediate g <0.65g. We found this trend was conserved across all spaceflight data sets, including both ecotypes of Arabidopsis and all three mutant lines. From the ImpulseDE2 results, we clearly see core trends in transcriptomic regulation not only in microgravity and Earth gravity, but at important intermediate gravity levels.

To evaluate the utility of partial gravity studies, the differential expression for the PGP-ESTEC dataset was compared to differential expression from spaceflight datasets. The PGP Flight, PGP-ESTEC, and SG2 datasets all include the Columbia ecotype, while SG1 and SG3 do not. To compare the effects of different types of gravity stimulation (true or simulated), the Columbia plants from these three datasets were compared. Microgravity and Martian gravity were selected to avoid spaceflight being a variable, as the SG2 dataset kept its 1g control on Earth. Data normalization and differential expression was calculated separately for each of the datasets and results were compared.

The spaceflight datasets consistently showed more similarity to other spaceflight data than they did to simulated gravity conditions. When quantifying the overlap in differentially expressed genes (DEGs) for microgravity to Martian gravity, the PGP-ESTEC transcriptome showed the lowest number of shared genes. At first, this difference might appear to be caused by fewer total differentially expressed genes identified in the PGP-ESTEC dataset. However, this does not explain why only 63% of the DEGs in the ESTEC data were shared, with other experiments while the other datasets averaged 81% gene overlap. These findings suggest that simulated gravity does not induce an equivalent response to true gravity.

From the differential expression, we could see the effects of spaceflight outside of gravity had a substantial impact on what genes were shared between datasets. The PGP Flight experiment grew plants under microgravity and 1g on the ISS. Seedling Growth 2 (SG2) grew plants in microgravity on the ISS but kept their 1g plants as a ground control on Earth. PGP-EMCS simulated 1g and microgravity on Earth using random positioning machines (RPMs) and RPM centrifuges. Because spaceflight is a variable between these datasets, but gravity is not, we can identify the effects of other spaceflight stressors while controlling for the effects of gravity.

The impact of radiation, fluid effects, and other spaceflight variables was quantified by the lack of overlap between datasets at 1g. The large overlap within experiment, compared to the small overlap between experiments, highlights the importance of other spaceflight variables. Each additional stress to the plant produces a compounding stress that can only be induced when all stresses of spaceflight are present.

To further identify which gene families were regulated in response to the stress of spaceflight, weighted gene co-expression was used to find gene clusters that shifted together between datasets. While this analysis did provide a list of potentially interesting genes, the main differences highlighted were across genotypes and RNA extraction types, as opposed to environmental change. Although these results are not immediately helpful in interpreting transcriptome responses to spaceflight, it does give insights into data approaches for open science. Not all these datasets used long reads, paired reads, or synthetic RNA spike-ins. Because of this inherent difference, general normalization techniques had to be used across fundamentally different extraction types for the weighted gene contrast analysis. These findings highlight the importance of setting technical standards for biological extractions on spaceflight tissues. The point in the extractions where spike-ins are added can also be a source of technical variation.

The variety of experimental designs used by primary investigators has provided a wealth of data for this meta-analysis. The differences in genotype, controls, and lighting provide evidence of conserved response pathways to the only consistent feature, gravity. While design differences have strengthened the analysis, consistency in technical approaches to RNA sequencing could add to the accessibility of future open data.