I’m conducting a metamodel study of the software called ‘LIFESIM’, but I don’t understand how to write the computational model in ‘UQLab’ because the code is not open source and the model is complex.

I’m trying to understand how to compare results from my computational model with the metamodel. If I input the same data and run it through the metamodel, wouldn’t I get the same results?

I think it will depend on the model you choose. For example, it is enough to use only the input-output dataset for a Kriging model. However, for a PCE model, I think you should also define the input object to create appropriate basis functions.

As discussed above, you can use your input/output data directly with certain surrogate models such as Kriging (a.k.a. Gaussian process models), support vector machines, etc. When using (sparse) polynomial chaos expansions, you need to define also an Input object that contain the distributions of the parameters which your input samples were drawn from, e.g. uniform, Gaussian, etc.
If you generated these samples by your self, you probably have this information at hand.

You will find a similar UQLab example here. Your data is simply passed to the metamodel options as: