Iterative motif search

[ozgunbabur]

Write an iterative method that will start by looking for enrichments and deficiencies for each location and each amino acid. Then it will

• select the significant subset using a given threshold
• for each significant result, do a new search within the pattern of that result

Iterate this until nothing comes out as significant. This will give you a tree of results, but you will see that some of the nodes on the tree will converge in the same motif, meaning it is actually a DAG.

Report each significant motif with their p-values, and with their parent-child relations.

[AdamFinkleUMB]

Tentatively done but not sure if result is what you want.

[ozgunbabur]

Hi Adam, please describe what you have done about this issue and please tell how we can test it.

[AdamFinkleUMB]

I fixed the bug we saw today: I needed to convert the number of the motif into a character. The search now neatly returns a readable result if the threshold is kept low.

[ozgunbabur]

What is the result on the simulated dataset with window 5?

[AdamFinkleUMB]

path = "test_data/simulated-phosphoproteomic-data.txt"
window = 5; length = 2 * window + 1
step = 1024
threshold = 0.0005

Key:
motif => [newfound_motifs]
(index, letter, presence)
Index is relative to 0 at the left, letter is the amino acid,
and presence is whether the acid must appear (True) or absent (False)

22212
None => [(4, 'S', False), (5, 'P', True)]

18981
(4, 'S', False) => [(5, 'P', True), (9, 'I', False)]

1239
(5, 'P', True) => []

15209
(9, 'I', False) => [(4, 'H', True), (5, 'P', True), (7, 'K', True)]

440
(4, 'H', True) => []

1017
(5, 'P', True) => []

1918
(7, 'K', True) => [(5, 'P', True)]

106
(5, 'P', True) => []

1512
(5, 'P', True) => []

Final Graph: {None: (5, 'P', True)}

[ozgunbabur]

How should we read these? I would like to understand the resulting DAG structure.

[AdamFinkleUMB]

The resulting acyclic graph in this case would be the original sequences with a single edge of (5, "P", True) leading to only the sequences with a "P" at index 5. The method itself works, and understanding the DAG structure can be part of the visualization issue.

[ozgunbabur]

Let me give an example: The output above produces "(5, 'P', True) => []" twice in the last steps. Why is that?

How can we look at this output and draw the DAG?

Also, where does the index 5 map on the sequence? Is it the center?