Reproducibility and Replication


After years of experience with E. coli arrays, including the GenoSys membranes, we can now routinely use them for robust gene expression profiling that properly reflects “biological reality”.  Replication is the key!  We routinely use 3 or more cultures for typical laboratory (flask or testtube) experiments – biological replicates.  We routinely obtain two technical replicates for each RNA sample (2 separate labeling/hybridizations, same RNA).  We treat the two spots on the array as independent measurements and normalize arrays by expressing the intensity of each spot as a percentage of the sum of all spot intensities.  In flask and testtube cultures, the correlation coefficients for all spots is usually >0.98 for technical replicates and >0.95 for biological replicates.  We occasionally obtain results that must be thrown out, or averaged with additional replicates.  The most dramatic improvement in culture-to-culture reproducibility came when we switched from flasks to a fermenter.  We now grow almost all of our batch cultures in a 1liter volume on a Braun BiostatB with constant conditions and pH-stat.  The biological reproducibility of technical replicates from different cultures by this method is always >0.99 (correl).

Top figure: Technical replicates of one RNA sample hybridized to a pair of “matched” membranes.

Bottom figure: Biological replicates of two cultures: E. coli MG1655 grown in fermenter on MOPS-glucose minimal medium at pH7.4, aerobically, 37oC.  Data represent average of 2 hybridizations.