Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
Model input often have to be estimated from experimental data as well. D2D offers several options when it comes to the estimation of inputs. When the input has to be estimated from noisy data, it is often a good idea to use a spline interpolant with relatively few knots. For a statistically sound description of the model such input signals should be comprehensively estimated with the model dynamics.
- Schelker M., et al. Comprehensive estimation of input signals and dynamical parameters in biochemical reaction networks. Bioinformatics 28(18), i522-i528, 2012.
When the input data contains strong transients, splines can have a tendency to over and undershoot between knots. In such cases, it may be better to use monotonic splines since they do not suffer from over and undershoot behaviour as much as the regular cubic splines. Monotonic splines are also useful for modelling input sources which are not very noisy, or inputs requiring a large number of knots. Setting the first and last two knot parameters to the same value guarantees that the monotonic spline remains constant outside the spline range. This means that monotonic splines can be combined into splines comprising of more than 10 knots (see Examples/LongSplines).
For implementational details and more information, see the section 1.5 under
Setting-up-models. One example of modeling an input using cubic splines is shown in the example application JAK/STAT signaling model. For an example comparing the spline types see
Examples/LongSplines. Note: invoking ar.config.turboSplines = 1, forces D2D to use a spline implementation which has higher performance, but forces recompilation of the conditions whenever the model is recompiled.