Normalization

[BenxiaHu]

Hello,
i just want to know whether HiCExplorer uses ICE normlaization (HiC-pro) or balanced (cooler)?

[joachimwolff]

You can use KR normalization or ICE and apply this with hicCorrectMatrix. If you have a cool matrix from somewhere else, we apply by default whatever is written in the weight column of the cool file. It doesn’t matter which correction method created them. bxhu ***@***.***> schrieb am Do. 19. Dez. 2024 um 14:32:

…

Hello, i just want to know whether HiCExplorer uses ICE normlaization (HiC-pro) or balanced (cooler)? — Reply to this email directly, view it on GitHub <#919>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADGQCADR2IYBEBU7H3KVIQ32GLDGVAVCNFSM6AAAAABT5AA7QGVHI2DSMVQWIX3LMV43ASLTON2WKOZSG42TAMZUGY4DSNY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[BenxiaHu]

@joachimwolff
i am a little confused the normlized matrix by Hicexploer compared to cooler.
the range of normlized matrix by Hicexploer is more than 1.
however, range of normlized matrix by cooler is less than 1.

huge difference.

[joachimwolff]

We scale back to the original value range because we have count values and it is nonsense in our opinion to have therefore values between 0 and 1. bxhu ***@***.***> schrieb am Do. 19. Dez. 2024 um 15:04:

…

@joachimwolff <https://github.com/joachimwolff> i am a little confused the normlized matrix by Hicexploer compared to cooler. the range of normlized matrix by Hicexploer is more than 1. however, range of normlized matrix by cooler is less than 1. huge difference. — Reply to this email directly, view it on GitHub <#919 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADGQCAATDC4LLS7YFIMDWU32GLG5BAVCNFSM6AAAAABT5AA7QGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNJUGE3DSMRZG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[BenxiaHu]

it is **nonsense** in our opinion to have therefore values between 0 and 1.--- that is true, and I also agree with you.

We scale back to the original value range because we have count values--- can you explain a little bit about the normalization algorithm (e.g., ICE)? thanks a lot.

[joachimwolff]

ICE is the algorithm as described by Imkaev 2012: https://pmc.ncbi.nlm.nih.gov/articles/PMC3816492/ You have the assumption that every region should interact with all other regions in sum by the same amount. That is tried to achieve by iterating over it and changing the values accordingly until the goal is achieved. I don’t have the formula in mind for each pixel entry, but that you can find in the linked paper. bxhu ***@***.***> schrieb am Do. 19. Dez. 2024 um 15:12:

…

it is **nonsense** in our opinion to have therefore values between 0 and 1.--- that is true, and I also agree with you. We scale back to the original value range because we have count values--- can you explain a little bit about the normalization algorithm (e.g., ICE)? thanks a lot. — Reply to this email directly, view it on GitHub <#919 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADGQCABAY625SBQ5FM2YSZD2GLH6XAVCNFSM6AAAAABT5AA7QGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKNJUGIYTQNJYHA> . You are receiving this because you were mentioned.Message ID: ***@***.***>