search

broadinstitute / ABC-Enhancer-Gene-Prediction

Cell type specific enhancer-gene predictions using ABC model (Fulco, Nasser et al, Nature Genetics 2019)
MIT License
199 stars 59 forks source link

Calculating ABC score using HiChIP and expression data #206

Closed anapkut closed 4 months ago

anapkut commented 5 months ago

Dear authors,

I am trying to run the ABC model using publicly available ATACseq, H3K27ac, H3K27ac HiChIP and my own expression data, but am running into a lot of issues and was wondering if you could help me. I have listed my questions below:

  1. Even though I set a gene expression cut-off of 30 in my config file, it still uses a cut-off of 1 to run the model (see parameters.predict.txt config.yaml.txt)
  2. I know that HiChIP data is suboptimal for the ABC model, but it is unfortunately all I could find for my system. Even though ABC processes my ATAC and H3K27ac correctly calling enhancers in the expected regions (see screenshot below), it uses a completely different interactome to perform region-gene association and to score enhancers (see below - output from attached cong.yaml and parameters.predict.txt files). I have played around with the gamma and scale parameters, but the interactome does not change, although the fitted powerlaw fit does fit the data points. Have you ever seen this? Any suggestions as to what I can do?

Thank you very much!

Best,

Ana Patricia Kutschat

image

atancoder commented 5 months ago
  1. promoter_activity_quantile_cutoff 0.3 is the one we use. expression_cutoff is for when gene expression data is provided (which isn't something supported in the codebase as of now).
  2. I would debug by looking at the EnhancerPredictionsAllPutative file and locating the E-G pairs that match your HiChIP data. You can look at the hic_contact column to see the score directly computed from your data. hic_contact_pl_scaled_adj is the actual value used for contact, which incorporates some powerlaw adjustments.
engreitz commented 5 months ago

I would not recommend using the HiChIP data at this time — we have not developed the codebase to enable this. I suggest running ABC using just your ATAC and H3K27ac data Best, Jesse

On Wed, Mar 20, 2024 at 11:54 AM Anthony Tan @.***> wrote:

  1. promoter_activity_quantile_cutoff 0.3 is the one we use. expression_cutoff is for when gene expression data is provided (which isn't something supported in the codebase as of now).
  2. I would debug by looking at the EnhancerPredictionsAllPutative file and locating the E-G pairs that match your HiChIP data. You can look at the hic_contact column to see the score directly computed from your data. hic_contact_pl_scaled_adj is the actual value used for contact, which incorporates some powerlaw adjustments.

— Reply to this email directly, view it on GitHub https://github.com/broadinstitute/ABC-Enhancer-Gene-Prediction/issues/206#issuecomment-2010368517, or unsubscribe https://github.com/notifications/unsubscribe-auth/AA3LGEFAG5WKGKMGY2FTXYLYZHLNFAVCNFSM6AAAAABE37IGTOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDAMJQGM3DQNJRG4 . You are receiving this because you are subscribed to this thread.Message ID: <broadinstitute/ABC-Enhancer-Gene-Prediction/issues/206/2010368517@ github.com>

-- Jesse Engreitz, PhD Assistant Professor, Department of Genetics BASE Research Initiative, Betty Irene Moore Children’s Heart Center Stanford University School of Medicine www.engreitzlab.org +1 (206) 310-3935 Pronouns: He/him/his