The general strategy was to unzip the data on the fly, convert to tabular format and then select a random 1 million sequences and then submit these to fastx_quality_stats. So having a look through the output file, shows very high quality scores with median scores >36 which suggests this dataset is very high quality. Below see the code used and a graph of median quality scores throughout the run.
for FQZ in *bz2
do echo $FQZ
pbzip2 -dc $FQZ | paste - - - - | shuf | head -1000000 \
| tr '\t' '\n' | fastx_quality_stats | cut -f1,6-9
done | tee quality_analysis.txt
|Mean cycle base quality for GSE55123 RNA-seq data shows very high quality sequence reads.|