Correlations between genomic GC levels and optimal growth temperatures are not 'robust'.