Preconditioned Metropolis sampling as a strategy to improve efficiency in Posterior exploration**The work was supported by the Swedish Research Council within the UPMARC Linnaeus center of Excellence (S. Engblom) and by BMBF (Germany) project PrevOp-OVERLOAD, grant number 01EC1408H (V. Sunkara).

In the low copy number regime, the dynamics of chemically reacting systems is accurately modeled as a continuous-time Markov chain and the associated probability density obeys the chemical master equation. Parameter inference in such models is very challenging for various reasons: large levels of noise implies that large amount of data is required for identification, the presence of transient phases may shadow subsets of the parameters, and accurate likelihood estimation requires the solutions to master equations. The latter is itself a computational very challenging problem and although many approximate computational methods have been proposed previously, the final implied accuracy in estimated rate parameters is difficult to assess.