Jimin Tan

Hi could you post your code so I could reproduce the error?

> Hello jimin > I noticed that you have uploaded the code for the insulation score. However, I have a small question. The values in the matrix should mostly be...

Hi @1944498970 I don't think standardization changes pearson/spearman correlations. I think the Pearson'r shown is calculated on unnormalized data.

Hi @kir1to455 Thank you for raising this issue! The move `feature_forward` function is actually for adjusting the difference between CNN vs Transformer. CNN will have channels/hiddens as the second feature...

Hi @kir1to455 , the decoder here is actually a dilated 2D conv resnet. It is very different from the typical transformer unless you consider ViT. I named it decoder because...

Hi @rebeccaronai , I'll describe the main idea here: 1. After predicting with C.Origami, you will have a 2mb by 2mb or 256 by 256 bin matrix. 2. For each...

Hi @archieandrews10 . Right now it would be challenging to apply to scATAC-seq since we need CTCF ChIP-seq as well. We are working on a version right now that can...

Hi @1944498970 the normalization might be different. If you want to re-train on already normalized data you can probably remove the Hi-C matrix log1p transformation in dataloader.

Hi @GMFranceschini, and @rebeccaronai the general idea is to predict multiple regions and merge them along the genome axis. The overlapping region should be averaged. The result will look like...

Yes, interactions beyond 2mb are large scale and usually represent compartments. You can probably predict them with some classic polymer-based models. C.Origami currently don't have the capacity at that level.