search

jaspershen-lab / ipop_aging

Code for aging project
MIT License
26 stars 11 forks source link

object_cross_section_loess and object_cross_section #4

Open simplefywoo opened 2 months ago

simplefywoo commented 2 months ago 
Hi, thank you very much for your valuable open source data. 
I tried to use the metabolomics matrix provided in the article to reproduce the Sankey diagram of Figure E in F1. At first I took the 3_DEG.R runtime program in 3_metabolomics, during which I used object_cross_section_loess@sample_info$sample_id to replace the non-existent adjusted_age data, and I ended up with inconsistent results. 
Then I used 3_DEG_feature.R to run the program, but I found out that peak/object_cross_section doesn't exist either, what exists is object_cross_section_loess, and in the end my results are still inconsistent. 
I think I didn't figure out the relationship between object_cross_section and object_cross_section_loess very well, and would like to ask which matrix was used for the final Sankey diagram data analysis.
simplefywoo commented 2 months ago

Cases such as this also exist in 3_fuzzy_c_means_clustering_cross_section_loess.R: object_cross_section_loess did not find the adjusted_age information needed for the plot. I used the number in sample_id instead, but the subsequent plotting function reported an error. str(object_cross_section_loess@sample_info) 'data.frame': 99 obs. of 3 variables. $ sample_id : chr “26” “26.5” “27” “27.5” ... $ class : chr “Subject” “Subject” “Subject” “Subject” ... $ subject_id : chr “Subject” “Subject” “Subject” “Subject” ...

jaspershen commented 2 months ago

Thank you. object_cross_section is the data from which we get the mean value for each molecule/microbe for each person. And object_cross_section_loess is the data on which we use the loess method to smooth the object_cross_section. You can find the details in the methods section of the paper. If you want to repeat panel e of Figure 1, you should use the object_cross_section.

adb3502 commented 2 months ago

Hi! Thank you very much for making the data and source code open-source! This is high-quality data! I am attempting to reproduce the transcriptomic plot in Fig 4 and I ran the loess_fit_cross_section_same_samples.R script on the object_cross_section before running DE-SWAN2. However, I am getting the attached figure which looks different than the original one. There are 2 peaks at 44 and 60 but the peak at 60 is much higher. Is there something I'm missing. Do any of the parameters in the loess function need to be changed? Thanks. Edit: On closer inspection, I think there is a difference in the real_age parameter which seems to be slightly different than adjusted_age. I can't seem to find this data however and I doubt if this could cause such high differences observed.. transcriptomics DE-SWAN

jaspershen commented 2 months ago

Thanks. We should use the adjusted_age. I am not sure which data you used. But in my analysis, we used all the molecules/microbes to run the DE-SWAN and then separated them into different omics.

simplefywoo commented 2 months ago

Thank you. object_cross_section is the data from which we get the mean value for each molecule/microbe for each person. And object_cross_section_loess is the data on which we use the loess method to smooth the object_cross_section. You can find the details in the methods section of the paper. If you want to repeat panel e of Figure 1, you should use the object_cross_section.

hi,I used object_cross_section to duplicate figure 1e but it fails after several more attempts. The program I am using is 1-code/3_metabolomics/3_DEG_feature.R and the data loaded is 3-data_analysis/plasma_metabolomics/data_preparation/metabolite/object_cross_section, but first the anova_p is different from 3-data_analysis/plasma_metabolomics/DEG/cross_section_loess/anova_p. The number of differential metabolites in each time period in the final Sankey diagram obtained is also different. I carefully read the methodology (Dysregulated molecules compared to baseline during human aging) section of the article, which states:Before the statistical analysis, all the confounders were corrected, could it be that the expression matrix in object_cross_section that I read is an uncorrected matrix? I used the lm_adjust function in 1-code/100-tools.R to correct the data frame and then use it directly as subject_data, but the anova_p I get is still different from the 3-data_analysis/plasma_metabolomics/DEG/crosssection loess/anova_p.

jaspershen commented 2 months ago

Thank you. We did this for different omics separately. For example, for transcriptomics, you can use this code to repeat figure 1e. https://github.com/jaspershen-lab/ipop_aging/blob/main/1-code/1_transcriptomics/3_DEG_permutation_pvalue.R

simplefywoo commented 1 month ago

Thank you. We did this for different omics separately. For example, for transcriptomics, you can use this code to repeat figure 1e. https://github.com/jaspershen-lab/ipop_aging/blob/main/1-code/1_transcriptomics/3_DEG_permutation_pvalue.R

Thank you very much for your patience and guidance! I still have a few confusing points though:

  1. https://github.com/jaspershen-lab/ipop_aging/blob/main/1-code/1_transcriptomics/3_DEG_permutation_pvalue.R mentioned “3-data_analysis/ plasma_transcriptome/data_preparation/object_cross_section_loess” I didn't find it, I only see object_cross_section.
  2. I found no adjusted_age information in object_cross_section_loess after using the fitting method provided in the article. Also I found that other data (e.g. metabolomics) provided object_cross_section_loess also does not have adjusted_age information, but this information is needed to run DEG_permutation_pvalue.
  3. Is the expression matrix in object_cross_section/object_cross_section_loess already corrected for confounders? Do I need to further correct for confounders in the expression matrix before going for downstream analysis? I'm guessing this may be why the marker_each_point_permutation I analyze is always different from the article. Thanks again for your answers and patience!
LiGuangming309 commented 1 month ago

Thank you. We did this for different omics separately. For example, for transcriptomics, you can use this code to repeat figure 1e. https://github.com/jaspershen-lab/ipop_aging/blob/main/1-code/1_transcriptomics/3_DEG_permutation_pvalue.R

Thank you very much for your patience and guidance! I still have a few confusing points though:

  1. https://github.com/jaspershen-lab/ipop_aging/blob/main/1-code/1_transcriptomics/3_DEG_permutation_pvalue.R mentioned “3-data_analysis/ plasma_transcriptome/data_preparation/object_cross_section_loess” I didn't find it, I only see object_cross_section.
  2. I found no adjusted_age information in object_cross_section_loess after using the fitting method provided in the article. Also I found that other data (e.g. metabolomics) provided object_cross_section_loess also does not have adjusted_age information, but this information is needed to run DEG_permutation_pvalue.
  3. Is the expression matrix in object_cross_section/object_cross_section_loess already corrected for confounders? Do I need to further correct for confounders in the expression matrix before going for downstream analysis? I'm guessing this may be why the marker_each_point_permutation I analyze is always different from the article. Thanks again for your answers and patience!

I have the same confusing.