search

zhengjh39 / HiC-SGL

GNU General Public License v3.0
3 stars 0 forks source link

Questions about HiC-SGL #1

Open dangdachang opened 11 months ago

dangdachang commented 11 months ago

Hello, I read your wonderful work HiC-SGL published in Briefings in Bioinformatics and I have some questions:

  1. Whether the inter-chromosome contacts are included when using PCA to obtain the initial cell embedding for each single cell. Should the interactions of all chromosomes (including interchromosomal interactions) be compiled into a genome-wide contact matrix and then flattened into a vector before PCA? Or, should the contact matrix of each chromosome be flattened into a vector, and then the results from different chromosomes be concatenated into a long vector?
  2. In link prediction task, can HiC-SGL predict whether there is interaction between two regions from different chromosomes?

Thanks a lot and I'm looking forward to your reply.

zhengjh39 commented 11 months ago

Thank you for interest in my research. It should be noted that the inter-chromosomal contacts exhibit significantly lower strength relative to the intra-chromosomal interactions. Consistent with prior studies in the field. In my analysis, I ignore the inter-chromosomal contacts. Moreover, during the generation of cell features, PCA is applied to each chromosome independently and we aggregate the embeddings of all chromosomes of one cell to obtain the cell embedding.

------------------ 原始邮件 ------------------ 发件人: "zhengjh39/HiC-SGL" @.>; 发送时间: 2023年12月25日(星期一) 下午4:51 @.>; @.***>; 主题: [zhengjh39/HiC-SGL] Questions about HiC-SGL (Issue #1)

Hello, I read your wonderful work HiC-SGL published in Briefings in Bioinformatics and I have some questions:

Whether the inter-chromosome contacts are included when using PCA to obtain the initial cell embedding for each single cell. Should the interactions of all chromosomes (including interchromosomal interactions) be compiled into a genome-wide contact matrix and then flattened into a vector before PCA? Or, should the contact matrix of each chromosome be flattened into a vector, and then the results from different chromosomes be concatenated into a long vector?

In link prediction task, can HiC-SGL predict whether there is interaction between two regions from different chromosomes?

Thanks a lot and I'm looking forward to your reply.

— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: @.***>