I would like to propose some changes to facilitate the incorporation of variability in experimental data.
While running a prediction, if the SDfile with test compounds contains a column that matches the field "SDFile experimental" in the parameters file of the selected model, the model returns the experimental value instead of generating a new prediction (for the compounds for which the data are present). With the current settings in Flame, the error associated with experimental values is not considered and thus the confidence interval width for these predictions is 0.
It would be very useful to add one more option to the model parameters, for instance called "SDFile experimental error", which would allow the users to define the level of error for each experimental value separately. In case the user provides a SDfile with experimental values but no error levels, the latter could be assigned a default value. This feature would be very valuable, especially in the context of uncertainty quantification and propagation.
I would like to propose some changes to facilitate the incorporation of variability in experimental data.
While running a prediction, if the SDfile with test compounds contains a column that matches the field "SDFile experimental" in the parameters file of the selected model, the model returns the experimental value instead of generating a new prediction (for the compounds for which the data are present). With the current settings in Flame, the error associated with experimental values is not considered and thus the confidence interval width for these predictions is 0.
It would be very useful to add one more option to the model parameters, for instance called "SDFile experimental error", which would allow the users to define the level of error for each experimental value separately. In case the user provides a SDfile with experimental values but no error levels, the latter could be assigned a default value. This feature would be very valuable, especially in the context of uncertainty quantification and propagation.