Metabolic flux data dimension reduction and clustering rationale

Joyjoyjoyc commented 5 months ago

Hi everyone, Here is a question for me, as the author performed cube root transformation on TCGA LUAD flux output to preserve the sign(direction) of fluxes. If it possible to use the absolute value of fluxes as an input? which means we only care about the metabolic activity but not the direction? Thank you so much!

LiuCanidk commented 5 months ago

Hi, I guess so. But across samples, if you do not take the direction (i.e., sign) into account, I'm afraid you may miss lots of important differences.

Joyjoyjoyc commented 5 months ago

Thank you for your reply!

Hi, I guess so. But across samples, if you do not take the direction (i.e., sign) into account, I'm afraid you may miss lots of important differences.

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

Joyjoyjoyc commented 3 months ago

Hi, I guess so. But across samples, if you do not take the direction (i.e., sign) into account, I'm afraid you may miss lots of important differences.

Hi @LiuCanidk, new question for me, I wonder if you can figure it out. In this paper the Fig. 5d given below use the mean glucose uptake score to make a comparison between different cell types, and I wonder if they average the absolute flux (from 100 bootstrap results) for one sample(patient) or still keep the flux sign? Thank you so much!

LiuCanidk commented 3 months ago

Yes, @Joyjoyjoyc I think the figure caption of Fig.5d illustrated this clearly: "A box plot of mean glucose uptake flux of each cell type and the whole TME for the seven patients." And I think it keeps the flux sign, and for basic glucose uptake flux, the sign may be (I forget the sign in my data) all positive, so no problem for integration of different signs (positive & negative) Hope it helps!

Joyjoyjoyc commented 3 months ago

Yes, @Joyjoyjoyc I think the figure caption of Fig.5d illustrated this clearly: "A box plot of mean glucose uptake flux of each cell type and the whole TME for the seven patients." And I think it keeps the flux sign, and for basic glucose uptake flux, the sign may be (I forget the sign in my data) all positive, so no problem for integration of different signs (positive & negative) Hope it helps!

Thanks for your quick reply! cause in my data there are both negative and positive fluxes, so I'm a little bit confused whether to keep the sign or not. As I'm more concerned about some specific reactions' activity but not the direction(like the nutrient uptake or release), so I'm not sure it's suitable to use the absolute value for my own analysis. Thanks!

LiuCanidk commented 3 months ago

Thanks for your quick reply! cause in my data there are both negative and positive fluxes, so I'm a little bit confused whether to keep the sign or not. As I'm more concerned about some specific reactions' activity but not the direction(like the nutrient uptake or release), so I'm not sure it's suitable to use the absolute value for my own analysis. Thanks!

As for your own data, I still recommend you to keep the sign. From my perspective, the sign difference is at the qualitative level and the absolute value difference is at the quatitative level. Both are important to find biological meanings, you can just draw them on a boxplot and keep the original data point on the plot, too.

Joyjoyjoyc commented 3 months ago

Thanks for your quick reply! cause in my data there are both negative and positive fluxes, so I'm a little bit confused whether to keep the sign or not. As I'm more concerned about some specific reactions' activity but not the direction(like the nutrient uptake or release), so I'm not sure it's suitable to use the absolute value for my own analysis. Thanks!

As for your own data, I still recommend you to keep the sign. From my perspective, the sign difference is at the qualitative level and the absolute value difference is at the quatitative level. Both are important to find biological meanings, you can just draw them on a boxplot and keep the original data point on the plot, too.

Got it, Thank you so much!

KChen-lab / METAFlux

Metabolic flux data dimension reduction and clustering rationale #9