Genotyping with DNAbased Decisiontree 2004 3 9 InHee

Genotyping with DNA-based Decision-tree 2004. 3. 9. In-Hee Lee

Point Mutation Data Site 1 Site 2 Site 3 Site 4 Site 5 Genotype C/G A/G C/C G/G C/T 1/2 C/C A/A C/T G/G C/C 1 / 10 B C/C A/A C/T G/A C/T 1 / 14 A C/C A/A C/C G/A C/T 1 / 14 B C/C A/G C/C G/G C/T 1 / 41 G/G C/C G/G T/T 2/2 G/C G/A C/T G/G T/C 2 / 10 B G/C G/A C/T G/A T/T 2 / 14 A G/C G/A C/C G/A T/T 2 / 14 B C/C G/G T/T 41 / 41 C/C A/G T/C G/G C/C 10 B / 41 C/C A/A T/T G/G C/C 10 B / 10 B Mutation type of one site => decision value

Decision Tree for Genotyping 1 st site C/C G/G C/G 2 nd site A/A A/G G/A G/G 3 rd site C/C C/T T/T C/C C/C C/T 4 th site G/A G/G G/G G/G G/A 5 th site 1/14 B 1/14 A 1/10 B 1/41 10 B/10 B 41/41 10 B/41 1/2 2/10 B 2/14 A

Decision Making with Decision Tree A B A Pre-defined or learned decision-tree A B Input data A A B A Decision making

Method 1 : How to Get Mutation Type? • For i-th site Site_i_A C/G Possible mutations at i-th site Site_i_B + C G Tag_i Site_i_A_1 Var_i Site_i_B_2 C C C G G C C Mutation Type: C/G G C G Mutation Type: G/G Longest one represents the mutation type of the site.

Method 1 : Decision Making • Link the fragments from each mutation site. – Four kinds of linkers. ¨Tag_i / Tag_i+1 ¨Site_i_B_2 / Tag_i+1 ¨Tag_i / Site_i+1_A_1 ¨Site_i_B_2 / Site_i+1_A_1 • Tag_i’s should be carefully designed so that the length can determine the genotype.

Method 2 : How to Get Mutation Type? • For i-th site Site_i_A C/G Possible mutations at i-th site Site_i_B + C C G For every possible type of mutations. (With varying Var_i’s) C G G Tag_i Var_i C C G G C Mutation Type: C/G G C C Mutation Type: G/G Longest one represents the mutation type of the site.

Method 2 : Decision Making • Link the fragments from each mutation site. – The lengths of Var_i’s should keep the order: ¨Var_i_1 Mut > Var_i_2 Mut ≥ Var_i_No. Mut – Only one type of linker. ¨Tag_i / Tag_i+1 • Tag_i’s should be carefully designed so that the length can determine the genotype.

Future Work • Test the scheme with normal genomic DNA. • Debug the scheme from experimental results!