DNA hybridization is fundamental to biotechnology applications like  sequencing, PCR, and DNA microarrays. The design of primers and  probes requires prediction of melting temperatures and optimum  structures for DNA duplexes or self-structure. Successful prediction  algorithms typically assume that hybridization thermodynamics depend  on nearest-neighbor base pairs. Our interest is in extending the  reach of quantitative thermodynamics to modified nucleic acids,  initially specifically Locked Nucleic Acid (LNA). LNA incorporation  increases duplex stability and confers nuclease resistance, and this  combination has made LNA popular in diverse applications requiring  short oligonucleotides. We have measured a large set of UV absorbance  melting curves to provide a database of nearest neighbor rules for  predicting optimal incorporation sites and sequences. The challenges  in this effort are first the design of  an affordable set of  sequences that still gives good coverage, then the extraction of  useful thermodynamics from noisy data, and finally analysis using  intuition and singular value decomposition to give the best  predictions using the smallest possible set of parameters. We have  also characterized all of the LNA-DNA mismatches as well as  positional variations. Rules of thumb and rules for algorithms will  be presented.