Different results between absoluted and doabsoluted? #26

xiasijian · 2023-02-24T02:51:09Z

The question is below

xiasijian · 2023-02-24T02:52:38Z

code is below
#####################################################
DoAbsolute(Seg = cnvkit_res, Maf = maf_file,
platform = "Illumina_WES",
copy.num.type = "total",
temp.dir="./temp/",
results.dir = "output_Doabsolute",
nThread = 1,
sigma.p = 0,
max.sigma.h = 0.2,
min.ploidy = 2,
max.ploidy = 10,
keepAllResult = FALSE,
primary.disease = "Gastric Cancer",
clean.temp = FALSE,
max.as.seg.count = 3000,
max.non.clonal = 0.05,
max.neg.genome = 0.005,
min.mut.af = 0.02,
min.no.mut = 5,
verbose = TRUE)
#########################################################
RunAbsolute(seg.dat.fn = paste0(outdir,patient,"_test_output.seg"),
maf.fn = paste0(outdir,patient,"_test_output.maf"),
platform = "Illumina_WES",
copy_num_type = "total",
sigma.p=0,
results.dir = "output_absolute",
sample.name=patient,
primary.disease= "Gastric Cancer",
max.sigma.h=0.2,
min.ploidy=2,
max.ploidy=10,
max.as.seg.count=3000,
max.neg.genome=0.005,
max.non.clonal=0.05,
verbose = TRUE,
min.mut.af=0.02)

xiasijian · 2023-02-24T02:55:04Z

ShixiangWang · 2023-02-24T07:23:55Z

Hi, @xiasijian

Thanks for your report about the comparison. As the DoAbsolute is just an open-source wrapper of ABSOLUTE, there are only two reasons for explaining the minor difference.

The used versions of ABSOLUTE are different.
Preprocessing did or default options used by the DoAbsolute.

I guess you are using the same version of ABSOLUTE, so the point 2 would be the only reason.

The code related to the reason has been given as the below.

DoAbsolute/R/DoAbsolute.R

Lines 18 to 55 in 8748cfb

    
           #' More detail about how to analyze ABSOLUTE results please see [this link](http://software.broadinstitute.org/cancer/software/genepattern/analyzing-absolute-data). 
        
           #' @param Seg a `data.frame` or a file (path) contains columns 
        
           #' "Sample", "Chromosome", "Start", "End", "Num_Probes", "Segment_Mean". 
        
           #' @param Maf MAF, default is `NULL`, can provided as `data.frame` or file path. 
        
           #' @param sigma.p Provisional value of excess sample level variance used for mode search. Default: 0 
        
           #' @param max.sigma.h Maximum value of excess sample level variance (Eq. 6). Default: 0.2 
        
           #' @param min.ploidy Minimum ploidy value to consider. Solutions implying lower ploidy values will be discarded. Default: 0.5 
        
           #' @param max.ploidy Maximum ploidy value to consider. Solutions implying greater ploidy values will be discarded. Default: 10 
        
           #' @param primary.disease Primary disease of the sample. Default: `NA` 
        
           #' @param platform one of "SNP_6.0", "Illumina_WES", "SNP_250K_STY". Default: "SNP_6.0" 
        
           #' @param temp.dir directory path used to store tempory files. Default: Absolute subdirectory under `tempdir()` 
        
           #' @param clean.temp if `TRUE`, auto-clean temp dir at the end. Default: `FALSE` 
        
           #' @param results.dir directory path used to store result files. Default: work directory 
        
           #' @param max.as.seg.count Maximum number of allelic segments. Samples with a higher segment count will be flagged as 'failed'. Default: 1500 
        
           #' @param max.non.clonal Maximum genome fraction that may be modeled as non-clonal (subclonal SCNA). Solutions implying greater values will be discarded. Default: 0.05 
        
           #' @param max.neg.genome Maximum genome fraction that may be modeled as non-clonal with copy-ratio below that of clonal homozygous deletion. Solutions implying greater values will be discarded. Default: 0.005 
        
           #' @param copy.num.type The type of copy number to be handled. Either total or allelic. Total is what this package for. Default: "total" 
        
           #' @param min.mut.af Minimum mutation allelic fraction. Mutations with lower allelic fractions will be filtered out before analysis. Default: 0.1 
        
           #' @param min.no.mut Minor allele frequency file, or NULL if one is not available. This specifies the data for somatic point mutations to be used by ABSOLUTE. Default: 5 
        
           #' @param verbose if `TRUE`, print extra info. Default: `FALSE` 
        
           #' @param nThread number of cores used for computation. Default: 1L 
        
           #' @param keepAllResult if `TRUE`, clean all results, otherwise clean result directory and keep most important results. Default: `TRUE` 
        
           #' @param recover if `TRUE`, recover previous unfinished work. 
        
           #' This is helpful when program stop unexpectedly when `clean.temp` is FALSE. Default: `FALSE` 
        
           #' @author Shixiang Wang <w_shixiang@163.com> 
        
           #' @return NULL 
        
           #' @import foreach doParallel data.table utils parallel 
        
           #' @export 
        
           #' @references Carter, Scott L., et al. "Absolute quantification of somatic DNA alterations in human cancer." Nature biotechnology 30.5 (2012): 413. 
        
           #' 
        
           DoAbsolute <- function(Seg, Maf = NULL, 
        
                                  sigma.p = 0, max.sigma.h = 0.2, min.ploidy = 0.5, max.ploidy = 10, 
        
                                  primary.disease = NA, platform = c("SNP_6.0", "Illumina_WES", "SNP_250K_STY"), 
        
                                  temp.dir = file.path(tempdir(), "Absolute"), clean.temp = FALSE, 
        
                                  results.dir = getwd(), max.as.seg.count = 1500, 
        
                                  max.non.clonal = 0.05, max.neg.genome = 0.005, copy.num.type = c("total", "allelic"), 
        
                                  min.mut.af = 0.1, min.no.mut = 5, verbose = FALSE, nThread = 1L, keepAllResult = TRUE, 
        
                                  recover = FALSE) {

DoAbsolute/R/DoAbsolute.R

Lines 136 to 141 in 8748cfb

    
           Seg$Chromosome <- as.character(Seg$Chromosome) 
        
           Seg$Chromosome <- gsub(pattern = "chr", replacement = "", Seg$Chromosome, ignore.case = TRUE) 
        
           Seg$Chromosome <- gsub(pattern = "X", replacement = "23", Seg$Chromosome, ignore.case = TRUE) 
        
           if (verbose) cat("-> Keeping only chr 1-23 for CNV data...\n") 
        
           autosome <- as.character(seq(1, 23)) 
        
           Seg <- Seg[Chromosome %in% autosome, ]

DoAbsolute/R/DoAbsolute.R

Line 203 in 8748cfb

maf <- maf[(t_alt_count / (t_ref_count + t_alt_count)) >= min.mut.af, ]

xiasijian · 2023-02-24T09:49:04Z

Thanks for your quick reply, but I found a problem in your example data.

xiasijian · 2023-02-24T09:50:45Z

当我抽取一定数量取跑的时候，测试数据在两种软件的纯度结果就有差异。如上图

xiasijian · 2023-02-24T09:51:20Z

然而，如果只是用原的测试数据，两者的结果是一样的。

ShixiangWang · 2023-02-24T10:35:15Z

Thanks.

xiasijian · 2023-02-24T10:44:32Z

I do not know why this is, can you tell me? Thanks.

ShixiangWang · 2023-02-24T10:58:41Z

然而，如果只是用原的测试数据，两者的结果是一样的。

我不是很理解，你不是说结果一致了吗？ 原的测试数据又是什么意思？

xiasijian · 2023-02-24T11:19:39Z

王老师，是这样的，我发现当MAF文件和Seg文件的数目比较少时(如上图，我做了一定抽样)，两者结果是不一样的，而我说的原测试数据，就是不进行任何抽样。

…

---Original--- From: "Shixiang Wang ***@***.***> Date: Fri, Feb 24, 2023 18:58 PM To: ***@***.***>; Cc: "Sijian ***@***.******@***.***>; Subject: Re: [ShixiangWang/DoAbsolute] Different results between absoluted anddoabsoluted? (Issue #26) 然而，如果只是用原的测试数据，两者的结果是一样的。我不是很理解，你不是说结果一致了吗？原的测试数据又是什么意思？ — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

ShixiangWang · 2023-02-24T11:27:08Z

那我只能理解为2次抽样的数据不一样。另外 DoAbsolute 看你使用并行没有，如果是的话不一定 set.seed 能 work。其他的可能的原因我上面也已经说了。我这边也不存在其他的可能解释。

一个最简单测试重复的办法是运行到调用 RunAbsolute 函数时，对比下这里的输入文件和你单独运行 ABSOLUTE 的输入文件是否一致。

DoAbsolute/R/DoAbsolute.R

Lines 285 to 295 in 8748cfb

    
           suppressWarnings(ABSOLUTE::RunAbsolute( 
        
             seg.dat.fn = seg_fn, maf.fn = maf_fn, 
        
             sample.name = samples[i], 
        
             sigma.p = sigma.p, max.sigma.h = max.sigma.h, 
        
             min.ploidy = min.ploidy, max.ploidy = max.ploidy, 
        
             primary.disease = primary.disease, platform = platform, 
        
             results.dir = cache.dir, max.as.seg.count = max.as.seg.count, 
        
             max.non.clonal = max.non.clonal, max.neg.genome = max.neg.genome, 
        
             copy_num_type = copy_num_type, 
        
             min.mut.af = min.mut.af, verbose = verbose 
        
           ))

你可以指定 cores = 1，然后运行 debug(ABSOLUTE::RunAbsolute)，然后跑 DoAbsolute，等待跳到运行上面代码之前。

xiasijian · 2023-02-24T12:01:11Z

王老师，这个输入是一致的。

…

---Original--- From: "Shixiang Wang ***@***.***> Date: Fri, Feb 24, 2023 19:27 PM To: ***@***.***>; Cc: "Sijian ***@***.******@***.***>; Subject: Re: [ShixiangWang/DoAbsolute] Different results between absoluted anddoabsoluted? (Issue #26) 那我只能理解为2次抽样的数据不一样。另外 DoAbsolute 看你使用并行没有，如果是的话不一定 set.seed 能 work。其他的可能的原因我上面也已经说了。我这边也不存在其他的可能解释。一个最简单测试重复的办法是运行到调用 RunAbsolute 函数时，对比下这里的输入文件和你单独运行 ABSOLUTE 的输入文件是否一致。 https://github.com/ShixiangWang/DoAbsolute/blob/8748cfb39eca5059f539a57829c49c83a869a31a/R/DoAbsolute.R#L285-L295 你可以指定 cores = 1，然后运行 debug(ABSOLUTE::RunAbsolute)，然后跑 DoAbsolute，等待跳到运行上面代码之前。 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

xiasijian · 2023-02-24T12:01:48Z

好的，谢谢，我试下您提到的方法。

…

---Original--- From: "Shixiang Wang ***@***.***> Date: Fri, Feb 24, 2023 19:27 PM To: ***@***.***>; Cc: "Sijian ***@***.******@***.***>; Subject: Re: [ShixiangWang/DoAbsolute] Different results between absoluted anddoabsoluted? (Issue #26) 那我只能理解为2次抽样的数据不一样。另外 DoAbsolute 看你使用并行没有，如果是的话不一定 set.seed 能 work。其他的可能的原因我上面也已经说了。我这边也不存在其他的可能解释。一个最简单测试重复的办法是运行到调用 RunAbsolute 函数时，对比下这里的输入文件和你单独运行 ABSOLUTE 的输入文件是否一致。 https://github.com/ShixiangWang/DoAbsolute/blob/8748cfb39eca5059f539a57829c49c83a869a31a/R/DoAbsolute.R#L285-L295 你可以指定 cores = 1，然后运行 debug(ABSOLUTE::RunAbsolute)，然后跑 DoAbsolute，等待跳到运行上面代码之前。 — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

xiasijian · 2023-02-25T03:56:25Z

01_Doabsolute_min.zip

王老师，这是我测试的脚本和输出，我载入了两者的最后保存的Rdata文件，确认输入的一样的，但是在model.res的结果却完全不一样，因此导致结果不一致。

xiasijian · 2023-02-25T04:18:36Z

王老师，还有一个问题，DoAbsolute最后的输出为NA，具体含义是？

ShixiangWang · 2023-02-25T07:49:32Z

01_Doabsolute_min.zip

王老师，这是我测试的脚本和输出，我载入了两者的最后保存的Rdata文件，确认输入的一样的，但是在model.res的结果却完全不一样，因此导致结果不一致。

那这就很奇怪了啊，你使用的应该是同一个absolute版本才对

ShixiangWang · 2023-02-25T07:50:45Z

那就是调用的默认参数不同，你再看看，不应该有其他解释，那太玄学了

xiasijian · 2023-02-25T08:47:50Z

王老师，确实很让人疑惑，我把您的脚本看了不下10遍，没发现很影响结果的代码，但当Seg文件和Maf文件的行数较少时，这两个软件就会出现偏差，这种偏差既体现在肿瘤纯度，也体现在最后得到的Rdata文件。

ShixiangWang · 2023-02-26T01:50:08Z

那就暂时不要纠结这个了，你直接用 ABSOLUTE 更好的话就直接用 for 循环调用即可，我这个包写的目的也就是为了方便一点。下面这段代码就给后人参考吧，我会在 README 中说明一下。你后续如果发现真正的原因也可以后续再讨论。

rm(list=ls())
gc()

library(DoAbsolute)
library(dplyr)
example_path = system.file("extdata", package = "DoAbsolute", mustWork = T)
library(data.table)
library(ABSOLUTE)

##set workspace
outdir="./output/"
setwd(workdir)

# segmentation file
seg_normal =  file.path(example_path, "SNP6_blood_normal.seg.txt")
seg_solid  =  file.path(example_path, "SNP6_solid_tumor.seg.txt")
seg_metastatic  = file.path(example_path, "SNP6_metastatic_tumor.seg.txt")

# MAF file
maf_solid  = file.path(example_path, "solid_tumor.maf.txt")
maf_metastatic  = file.path(example_path, "metastatic_tumor.maf.txt")

# read data
seg_normal = fread(seg_normal)
seg_solid = fread(seg_solid)
seg_metastatic = fread(seg_metastatic)
maf_solid = fread(maf_solid)
maf_metastatic = fread(maf_metastatic)

# merge data
Seg = Reduce(rbind, list(seg_normal, seg_solid, seg_metastatic))
Maf = Reduce(rbind, list(maf_solid, maf_metastatic))

Seg$Sample = substr(Seg$Sample, 1, 15)
Maf$Tumor_Sample_Barcode = substr(Maf$Tumor_Sample_Barcode, 1, 15)

test_Seg=Seg %>% subset(Sample %in% c("TCGA-DK-A1A6-01"))
test_Seg=test_Seg[1:42,]
table(test_Seg$Chromosome)
test_Maf=Maf %>% subset(Tumor_Sample_Barcode %in% c("TCGA-DK-A1A6-01"))
table(test_Maf$Chromosome)
test_Maf$Chromosome=ifelse(test_Maf$Chromosome=="X",23,test_Maf$Chromosome)
test_Maf=test_Maf[1:28,]



############################ DoAbsolute #################################
# test function
DoAbsolute(Seg = test_Seg, Maf = test_Maf, 
           platform = "SNP_6.0", 
           copy.num.type = "total",
           results.dir = "output_doabsolute",
           nThread = 1, 
           sigma.p = 0,
           max.sigma.h = 0.2,
           min.ploidy = 0.5,
           max.ploidy = 10,
           keepAllResult = FALSE, 
           primary.disease = "Tumor",
           clean.temp = FALSE,
           max.as.seg.count = 1500,
           max.non.clonal = 0.05,
           max.neg.genome = 0.005,
           min.mut.af = 0.1,
           min.no.mut = 0,
           verbose = TRUE)

############################ Absolute #################################
for(patient in c("TCGA-DK-A1A6-01")){
  print(patient)
  test_seg=test_Seg%>%subset(Sample==patient)
  test_maf=test_Maf %>% subset(Tumor_Sample_Barcode==patient)
  write.table(test_seg,file = paste0(outdir,patient,"_test_output.seg"),sep="\t",row.names = F,quote = F)
  write.table(test_maf,file = paste0(outdir,patient,"_test_output.maf"),sep="\t",row.names = F,quote = F)
  RunAbsolute(seg.dat.fn = paste0(outdir,patient,"_test_output.seg"),
              maf.fn = paste0(outdir,patient,"_test_output.maf"), 
              platform = "SNP_6.0",
              copy_num_type = "total",
              sigma.p=0, 
              results.dir = "output_absolute",
              sample.name=patient,
              primary.disease="Tumor",
              max.sigma.h=0.5,
              min.ploidy=0.5,
              max.ploidy=10,
              max.as.seg.count=1600,
              max.neg.genome=0.005, 
              max.non.clonal=0.05,
              verbose = TRUE, 
              min.mut.af=0.1)
  
}


######################################
## merge absolute data
######################################
ab_outdir="./output_absolute/"
all_files=list.files(paste0(ab_outdir))
rdata_files=all_files[grepl(pattern = ".RData",all_files)]
absolute.files=paste0(ab_outdir,rdata_files)
results.dir <- file.path(ab_outdir, "abs_summary")
CreateReviewObject("DRAWS_summary", absolute.files, results.dir, "total", verbose=TRUE)

load(file = paste0(results.dir,"/","DRAWS_summary.PP-modes.data.Rdata"))
calls.path = file.path(paste0(results.dir, "/","DRAWS_summary.PP-calls_tab.txt"))
modes.path = file.path(paste0(results.dir, "/", "DRAWS_summary.PP-modes.data.RData"))
output.path = file.path(paste0(outdir, "abs_extract"))
ExtractReviewedResults(calls.path, "test", modes.path, output.path, "absolute", "total")

xiasijian · 2023-02-26T07:28:55Z

好的，谢谢您，王老师🙏

…

---Original--- From: "Shixiang Wang ***@***.***> Date: Sun, Feb 26, 2023 09:50 AM To: ***@***.***>; Cc: "Sijian ***@***.******@***.***>; Subject: Re: [ShixiangWang/DoAbsolute] Different results between absoluted anddoabsoluted? (Issue #26) 那就暂时不要纠结这个了，你直接用 ABSOLUTE 更好的话就直接用 for 循环调用即可，我这个包写的目的也就是为了方便一点。下面这段代码就给后人参考吧，我会在 README 中说明一下。你后续如果发现真正的原因也可以后续再讨论。 rm(list=ls()) gc() library(DoAbsolute) library(dplyr) example_path = system.file("extdata", package = "DoAbsolute", mustWork = T) library(data.table) library(ABSOLUTE) ##set workspace outdir="./output/" setwd(workdir) # segmentation file seg_normal = file.path(example_path, "SNP6_blood_normal.seg.txt") seg_solid = file.path(example_path, "SNP6_solid_tumor.seg.txt") seg_metastatic = file.path(example_path, "SNP6_metastatic_tumor.seg.txt") # MAF file maf_solid = file.path(example_path, "solid_tumor.maf.txt") maf_metastatic = file.path(example_path, "metastatic_tumor.maf.txt") # read data seg_normal = fread(seg_normal) seg_solid = fread(seg_solid) seg_metastatic = fread(seg_metastatic) maf_solid = fread(maf_solid) maf_metastatic = fread(maf_metastatic) # merge data Seg = Reduce(rbind, list(seg_normal, seg_solid, seg_metastatic)) Maf = Reduce(rbind, list(maf_solid, maf_metastatic)) Seg$Sample = substr(Seg$Sample, 1, 15) Maf$Tumor_Sample_Barcode = substr(Maf$Tumor_Sample_Barcode, 1, 15) test_Seg=Seg %>% subset(Sample %in% c("TCGA-DK-A1A6-01")) test_Seg=test_Seg[1:42,] table(test_Seg$Chromosome) test_Maf=Maf %>% subset(Tumor_Sample_Barcode %in% c("TCGA-DK-A1A6-01")) table(test_Maf$Chromosome) test_Maf$Chromosome=ifelse(test_Maf$Chromosome=="X",23,test_Maf$Chromosome) test_Maf=test_Maf[1:28,] ############################ DoAbsolute ################################# # test function DoAbsolute(Seg = test_Seg, Maf = test_Maf, platform = "SNP_6.0", copy.num.type = "total", results.dir = "output_doabsolute", nThread = 1, sigma.p = 0, max.sigma.h = 0.2, min.ploidy = 0.5, max.ploidy = 10, keepAllResult = FALSE, primary.disease = "Tumor", clean.temp = FALSE, max.as.seg.count = 1500, max.non.clonal = 0.05, max.neg.genome = 0.005, min.mut.af = 0.1, min.no.mut = 0, verbose = TRUE) ############################ Absolute ################################# for(patient in c("TCGA-DK-A1A6-01")){ print(patient) test_seg=test_Seg%>%subset(Sample==patient) test_maf=test_Maf %>% subset(Tumor_Sample_Barcode==patient) write.table(test_seg,file = paste0(outdir,patient,"_test_output.seg"),sep="\t",row.names = F,quote = F) write.table(test_maf,file = paste0(outdir,patient,"_test_output.maf"),sep="\t",row.names = F,quote = F) RunAbsolute(seg.dat.fn = paste0(outdir,patient,"_test_output.seg"), maf.fn = paste0(outdir,patient,"_test_output.maf"), platform = "SNP_6.0", copy_num_type = "total", sigma.p=0, results.dir = "output_absolute", sample.name=patient, primary.disease="Tumor", max.sigma.h=0.5, min.ploidy=0.5, max.ploidy=10, max.as.seg.count=1600, max.neg.genome=0.005, max.non.clonal=0.05, verbose = TRUE, min.mut.af=0.1) } ###################################### ## merge absolute data ###################################### ab_outdir="./output_absolute/" all_files=list.files(paste0(ab_outdir)) rdata_files=all_files[grepl(pattern = ".RData",all_files)] absolute.files=paste0(ab_outdir,rdata_files) results.dir <- file.path(ab_outdir, "abs_summary") CreateReviewObject("DRAWS_summary", absolute.files, results.dir, "total", verbose=TRUE) load(file = paste0(results.dir,"/","DRAWS_summary.PP-modes.data.Rdata")) calls.path = file.path(paste0(results.dir, "/","DRAWS_summary.PP-calls_tab.txt")) modes.path = file.path(paste0(results.dir, "/", "DRAWS_summary.PP-modes.data.RData")) output.path = file.path(paste0(outdir, "abs_extract")) ExtractReviewedResults(calls.path, "test", modes.path, output.path, "absolute", "total") — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: ***@***.***>

xiasijian changed the title ~~Difference~~ Different results between absoluted and doabsoluted? Feb 24, 2023

ShixiangWang closed this as completed Feb 24, 2023

h170607 mentioned this issue Aug 30, 2023

estimate tumour purity and ploidy xiasijian/CNV_facets#2

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Different results between absoluted and doabsoluted? #26

Different results between absoluted and doabsoluted? #26

xiasijian commented Feb 24, 2023 •

edited

Loading

xiasijian commented Feb 24, 2023 •

edited

Loading

xiasijian commented Feb 24, 2023

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023

xiasijian commented Feb 24, 2023

xiasijian commented Feb 24, 2023

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023 via email

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023 via email

xiasijian commented Feb 24, 2023 via email

xiasijian commented Feb 25, 2023

xiasijian commented Feb 25, 2023

ShixiangWang commented Feb 25, 2023

ShixiangWang commented Feb 25, 2023

xiasijian commented Feb 25, 2023

ShixiangWang commented Feb 26, 2023

xiasijian commented Feb 26, 2023 via email

Different results between absoluted and doabsoluted? #26

Different results between absoluted and doabsoluted? #26

Comments

xiasijian commented Feb 24, 2023 • edited Loading

xiasijian commented Feb 24, 2023 • edited Loading

xiasijian commented Feb 24, 2023

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023

xiasijian commented Feb 24, 2023

xiasijian commented Feb 24, 2023

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023 via email

ShixiangWang commented Feb 24, 2023

xiasijian commented Feb 24, 2023 via email

xiasijian commented Feb 24, 2023 via email

xiasijian commented Feb 25, 2023

xiasijian commented Feb 25, 2023

ShixiangWang commented Feb 25, 2023

ShixiangWang commented Feb 25, 2023

xiasijian commented Feb 25, 2023

ShixiangWang commented Feb 26, 2023

xiasijian commented Feb 26, 2023 via email

xiasijian commented Feb 24, 2023 •

edited

Loading

xiasijian commented Feb 24, 2023 •

edited

Loading