To extend our previous fitting results [ 20], two in vitro kinetic datasets were fitted to models of the yeast Grx system (Table 1, [ 26]) and in both cases, good fits of the data were obtained (r≥0.94, Figure 3).
To characterise the circadian profile of the activity-rest cycle, individual actigraphy datasets were fitted to an extended Cosinor model using GraphPad Software.
As expected, in the case when the datasets were fitted using the analytical expressions with the same inter-division time distribution as used in generating them (gamma-distribution with shape parameter 3), we were able to obtain a good fit and recover the parameters used to simulate the data (results not shown).
All datasets were fitted with the same set of rate constants and SAS (except for one rate, see below).
Datasets were fitted either to a single exponential or, where single exponential fits were not satisfactory, to a double exponential function, as described previously [ 13, 14].
The two protein datasets were fitted into Normal distribution models and the threshold for significance of +/− 1.5 was sufficient to separate differentially expressed proteins.
To reduce the complexity of the model, in this last case, we made subsets of samples belonging to each cultivar and estimated the parameters separately.> -wrap-foot> Datasets were fitted to the corresponding formula (C = cultivar, Ti = time point, Tr = treatment; additive effects are represented by "+"; ":" represents interaction).
For both the shotgun proteomes and transcriptomes, replicate datasets were fitted with a locally weighted scatterplot smoothing by polynomial regression using a standard span value of 0.75 (R Core Development Team, 2014), using the replicates of unprocessed samples, and the replicate time series of 7.8 72 hr, as input for the regression.
> -wrap-foot> A key functionality of PW is multi-experiment fitting, where several datasets are fitted simultaneously.
