fisher test for significance of one type of transposable element

Question

[yasjas]

Hello everyone,

Basically I have two files (healthy VS cancer tissu) containing TE (transposable) elements and their sub-classes.

After I have done a counting for one type of class LTR for example in the two files separately, found out that their were more present in cancer than in healthy tissue, so my question is:

• How can I do a fisher test to say that maybe the presence of LTR is related to cancer, whether it is significant or not?

Thank you for any answer or thought I can get from you.

Jas

E sub_classes     TE classes
1                HAL1           LINE
2                HAL1           LINE
3                HAL1           LINE
4                L1M4           LINE
5            (CACCC)n  Simple_repeat
6                L1M4           LINE
7                MSTD            LTR
8               AluYc           SINE
9             GC_rich Low_complexity
10            GC_rich Low_complexity
11            GC_rich Low_complexity
12            GC_rich Low_complexity
13                L2a           LINE
14           (CCCCG)n  Simple_repeat
15            GC_rich Low_complexity
16            (CCCA)n  Simple_repeat
17            GC_rich Low_complexity
18             MLT1G1            LTR
19              MLT1H            LTR
20              AluJb           SINE
21              MLT1C            LTR

Answer 1

[Devon Ryan]

The four numbers you would need for a fishers test are:

• number of genes with TEs in healthy tissue
• number of genes without TEs in healthy tissue
• number of genes with TEs in cancerous tissue
• number of genes without TEs in cancerous tissue.

You can then make a matrix of that. You could alternatively count the number of TEs in genes or not in genes in both datasets and use that. Both are fine, but answer slightly different questions (the latter method is much less interesting if there's just a huge increase in TEs in one gene...).

[yasjas]

Thank you Devon