Recombinant protein production is a widely used technique, yet half of these experiments fail at the expression phase and only a quarter of target proteins are successfully purified. Failures are largely due to toxicity, low protein expression, and poor protein solubility. We have discovered that the energetics of RNA structure ensembles, that model the 'accessibility' of translation initiation sites, accurately predicts the expression outcomes of 11,430 recombinant protein production experiments in Escherichia coli. We have further discovered that normalised B-factors, that model the 'flexibility' of amino acid residues, accurately predicts the solubility of 12,158 recombinant proteins expressed in Escherichia coli. One important implication of our findings is that the chances of successful experiments can be increased by improving both 'accessibility' and 'flexibility'. We have developed TIsigner (Translation Initiation coding region designer) and SoDoPE (Soluble Domain for Protein Expression) that allows users to choose a protein region of interest for optimising expression and solubility, respectively. The final results will suggest synonymous codon changes within the first few codons of the DNA fragments of interest, meaning that gene optimisation can be done using standard PCR cloning.