Approach to design of DNA chips using self organizing map

Abstract

Recently almost the entire sequence of the human genome was determined. Now, genome function analysis is being actively investigated. DNA chips are useful tools for sequencing and SNP analyses. DNA chips consist of surfaces on which probes of DNA sub-sequences of short length (4-10b) and all combinations of DNA of a specific length are printed without any particular design. To determine DNA sequences longer than 1000 bases, a DNA chip with probes longer than 10 bases is desirable, but this may be result in a huge and prohibitively expensive chip. Therefore, to reduce the size of DNA chips, chip design is considered to be important. We applied Self Organizing Maps (SOM) to select a adequate set of probes to represent actual DNA sequences on a DNA chip. SOM is particularly well suited because it can organize the generic feature of DNA sequence by learning DNA sequences. Furthermore, it will be possible to design the 2-dimentional layout of the probes on DNA chip. We improved the SOM algorithm so that it could learn the discrete values for 'A','T','G',C' symbols and adjust the length of probes automatically. We examined the selected probes by testing for DNA polymorphism on a computer simulation to evaluate the adequacy of the selected sets of probes.