We have been using SecreC - a new programming language introduced in Sharemind 1.9 - for creating actual data mining applications. Our first attempts have been documented in a research report.
Bogdanov, D., Jagomägis, R., Laur, S. Privacy-preserving Histogram Computation and Frequent Itemset Mining with Sharemind. Cybernetica research report T-4-8. 2009.
Get the paper (T-4-8) from here.
The report describes how we adapted the algorithms for Sharemind. We also give a thorough analysis of the benchmark results. If you want to reproduce the results or see the algorithms in action, get Sharemind 1.91.