I would like a rough estimate for a free energy of substitution between different nucleotide bases e.g., from a purine to a pyrimidine, against its complement, with neighbouring bases taken into consideration. 

Only a very rough estimate. I want to know given a sequence, how likely is a single nucleotide polymorphism to occur, using thermodynamic principles.

As an example, this is a link to a tool that estimates the DeltaG of a helical protein sequence inserting into a lipid bilayer:

http://dgpred.cbr.su.se/