We compared the results of T test on beta values with chi-square test on SMP genotypes (MM vs. MU vs. UU, 2×3 contingency table) using a case-control data set GSE49031 (including 797 pediatric acute lymphoblastic leukemia patients and 148 normal controls, total 482421 DNA methylation loci). We chose several threshold levels, such as 0.01, 0.001, 1e-5, to observe the consistency of the results identified by T test on beta values and chi-square test on SMP genotypes.
Table 1-4 showed that the consistency is increasing with the increase of Fold Change (FC, the average beta value of one group divided by the average of the second groups) between two sample groups. All the results can be found as following:
Table1. fold_change>1.5
threshold levels |
results of T test (n) |
results of chi-square test (n) |
overlap |
consistency |
0.05 |
76889 |
76575 |
76201 |
0.993 |
0.01 |
74322 |
75515 |
72962 |
0.974 |
0.001 |
70515 |
72665 |
67634 |
0.945 |
1e-4 |
68264 |
69040 |
63674 |
0.928 |
1e-5 |
66473 |
62301 |
60054 |
0.934 |
Table2. fold_change>2 |
threshold levels |
results of T test (n) |
results of chi-square test (n) |
overlap |
consistency |
0.05 |
48963 |
48816 |
48816 |
0.998 |
0.01 |
48963 |
48625 |
48625 |
0.997 |
0.001 |
48916 |
47734 |
47723 |
0.988 |
1e-4 |
48550 |
46099 |
46012 |
0.973 |
1e-5 |
47998 |
43998 |
43870 |
0.956 |
Table3. fold_change>2.5 |
threshold levels |
results of T test (n) |
results of chi-square test (n) |
overlap |
consistency |
0.05 |
35586 |
35557 |
35557 |
0.9996 |
0.01 |
35586 |
35521 |
35521 |
0.9991 |
0.001 |
35586 |
35334 |
35334 |
0.9965 |
1e-4 |
35586 |
34796 |
34796 |
0.9889 |
1e-5 |
35578 |
33730 |
33730 |
0.9740 |
Table4. fold_change>3 |
threshold levels |
results of T test (n) |
results of chi-square test (n) |
overlap |
consistency |
0.05 |
26542 |
26540 |
26540 |
0.99996 |
0.01 |
26542 |
26532 |
26532 |
0.99981 |
0.001 |
26542 |
26505 |
26505 |
0.99930 |
1e-4 |
26542 |
26386 |
26386 |
0.99706 |
1e-5 |
26542 |
26013 |
26013 |
0.99003 |
supplementary doc file: supplementary.doc |