It depends on which kind of analysis you do. if your outputs are simple data representation such as .bed or .bedgraph files the liftOver tool from UCSC can help keep everything up to date. it might require more work in generating the chain files if you work with unusual organisms. this method works for almost any form of feature-based annotation (or even .wig files), the only thing you need is to convert it into a .bed or .bedgraph.

A brief word on how liftOver works. the program is designed to convert a set of genomic coordinates (with respective values if present) between different assemblies of the same organism. in order to do this, liftOver requires a special "chain file", a file containing the differencese between the two target assemblies.

a number of pre-computed chainfiles for a number of different organisms, along with liftOver itself, is downloadable here.

For UniProtKB, you could explore the advanced query mechanisms by date and use them programmatically (cf http://www.uniprot.org/faq/28 and http://www.uniprot.org/help/query-fields ), e.g.

lwp-mirror "http://www.uniprot.org/uniprot/?query=reviewed:yes+AND+created:[current+TO+current]&format=txt" new_seq.dat
lwp-mirror "http://www.uniprot.org/uniprot/?query=reviewed:yes+AND+sequence_modified:[current+TO+current]&format=txt" upd_seq.dat
lwp-mirror "http://www.uniprot.org/uniprot/?query=reviewed:yes+AND+modified:[current+TO+current]&format=txt" upd_ann.dat