search

loosolab / TF-COMB

Transcription Factor Co-Occurrence using Market Basket analysis
https://tf-comb.readthedocs.io
MIT License
10 stars 1 forks source link

Please kindly explain this figure #74

Closed Rajashree93 closed 7 months ago

Rajashree93 commented 9 months ago

Can you please tell me?

  1. Why HOXB6 is only showing changes while LX5 is not?
  2. How are you calculating the bigwig score? What is the biological implication of the bigwig score?
  3. What does the changing blue red (bigwig score based) inside the figure mean?
  4. In this particular figure, based on the binding strand colours, does it mean LX5 and HOXB6 are binding in two opposite strands?
  5. What does the -50 to 150 mean in the x axis?

GC_HOXB6-LHX5.pdf

hschult commented 8 months ago

Hi @Rajashree93

sure I will gladly explain this plot. We call this plot a pairMap it is a heatmap that shows all binding pairs of two transcription factors (TFs) sorted by the distance between them. The plot is intended to see if the TF-pair shows a preferred binding distance and possibly strand orientation.

Here is the example and description we showed in our paper: grafik Footprinting heatmap of all co-occurring JUNB-FOS sites. Colored for Tn5 cutsites appearing more than expected (red) or less than expected (blue) if DNA is inaccessible. Black lines represent the edges of JUNB (left) and FOS (right) motifs. Edges show the binding strand of the respective TF.

Now to your specific pairMap.

  1. I assume you are referring to the grey areas. These are the motif positions of the two TFs for each co-occurrence event (row). The position does not change for LHX5 since the plot is left-aligned. You can also create a centered or right-aligned view if you prefer (see pairMap align parameter).
  2. & 3. In the example bigwig score refers to the Tn5 cutsite abundance as described above which is a scoring provided by our TOBIAS tool. But it is possible to provide any bigwig file so I can not tell you what the scores mean in your case.
  3. You can not say that the TFs always occur on opposite strands but there are some distances where they seem to prefer opposite strands. For example at the bottom of your pairMap with a distance of ~100bp LHX5 seems to bind mostly on - whereas HOXB6 prefers the + strand.
  4. The x-axis is the relative distance to the center point in basepairs. In your case 0 is the right border of the LHX5 motif due to the left alignment you have chosen.

Thanks for your interest I hope that answers all of your questions.

Rajashree93 commented 8 months ago

Hi, Thank you. Do you know how to determine the p value of changing co-occurrence pairs? For example, I had a TF pair in control which has .58 cosine and in treatment with .63. The change in cosine is .5, how do I determine the significance/p-value for it?

Best, Rajashree

On Mon, Jan 8, 2024 at 8:57 AM hschult @.***> wrote:

Hi @Rajashree93 https://github.com/Rajashree93

sure I will gladly explain this plot. We call this plot a pairMap it is a heatmap that shows all binding pairs of two transcription factors (TFs) sorted by the distance between them. The plot is intended to see if the TF-pair shows a preferred binding distance and possibly strand orientation.

Here is the example and description we showed in our paper: grafik.png (view on web) https://github.com/loosolab/TF-COMB/assets/27274175/70960c1b-4426-4ee9-9762-031d71ad7189 Footprinting heatmap of all co-occurring JUNB-FOS sites. Colored for Tn5 cutsites appearing more than expected (red) or less than expected (blue) if DNA is inaccessible. Black lines represent the edges of JUNB (left) and FOS (right) motifs. Edges show the binding strand of the respective TF.

Now to your specific pairMap.

  1. I assume you are referring to the grey areas. These are the motif positions of the two TFs for each co-occurrence event (row). The position does not change for LHX5 since the plot is left-aligned. You can also create a centered or right-aligned view if you prefer (see pairMap https://tf-comb.readthedocs.io/en/latest/API/tfcomb.utils.html#tfcomb.utils.TFBSPairList.pairMap align parameter).
  2. & 3. In the example bigwig score refers to the Tn5 cutsite abundance as described above which is a scoring provided by our TOBIAS tool. But it is possible to provide any bigwig file so I can not tell you what the scores mean in your case.
  3. You can not say that the TFs always occur on opposite strands but there are some distances where they seem to prefer opposite strands. For example at the bottom of your pairMap with a distance of ~100bp LHX5 seems to bind mostly on - whereas HOXB6 prefers the + strand.
  4. The x-axis is the relative distance to the center point in basepairs. In your case 0 is the right border of the LHX5 motif due to the left alignment you have chosen.

Thanks for your interest I hope that answers all of your questions.

— Reply to this email directly, view it on GitHub https://github.com/loosolab/TF-COMB/issues/74#issuecomment-1881171198, or unsubscribe https://github.com/notifications/unsubscribe-auth/A67AXAANH7K67Z573ZEDED3YNQCL3AVCNFSM6AAAAABBD6CHVOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQOBRGE3TCMJZHA . You are receiving this because you were mentioned.Message ID: @.***>

hschult commented 7 months ago

Hi @Rajashree93!

Unfortunately, computing the p-value is not possible with this kind of data hence, why we implemented the z-score as an alternative. Also, please don't ask the same question multiple times. My colleague already answered this in #67.

I'm closing this issue since the original question was answered.