Questions about including covariates in the model

[xiw588]

Hi, I am trying to run multivariable survival analysis using the code but got the error message below. Any idea of why this is happening? Thanks!

surv_GIA<-coxph(Surv(age_diag, diagnosis_status) ~ term+Predicted_Pop+sex+smoking, data =HLA)

surv_GIA_supertype_uncond <- runMiDAS(
object = surv_GIA,
experiment = "hla_supertypes",
inheritance_model = "additive",
lower_frequency_cutoff = 0.05,
upper_frequency_cutoff = 0.98,
correction = "bonferroni",
exponentiate = TRUE
) %>% filter(p.adjusted<0.05)

---

Warning message:
While evaluating the model problems occured:

• Warnings:
• Using formula(x) is deprecated when x is a character vector of length > 1.
Consider formula(paste(x, collapse = " ")) instead.; occured 33 times, including variables: A01:01, A02:01, A03:01, A24:02, B*18:01, ...

[mcjmigdal]

Could you provide some minimal example that I could reproduce? If I were to guess you might need to eg. one hot encode your covariates so they are compatible with cox regression.

[xiw588]

Thank you for your response! What do you mean by one hot encode? There are continuous and categorical variables in the model. For categorical ones, they are in characters.

[mcjmigdal]

Without a reproducible example all I can do is guessing 🫣

[xiw588]

Thanks! I am attaching a sample dataset and the code that I used. phenotype_data<-as.data.frame(test_phenotype_0331)
hla_midas_alleles<-as.data.frame(test_0331)
HLA <- prepareMiDAS(
hla_calls = hla_midas_alleles,
colData = phenotype_data,
experiment = c('hla_alleles','hla_supertypes','hla_het','hla_divergence')
)
surv_GIA<-coxph(Surv(age_diag, diagnosis_status) ~ term+Predicted_Pop+sex+smoking, data =HLA)
surv_GIA_allele_uncond <- runMiDAS(
object = surv_GIA,
experiment = "hla_alleles",
inheritance_model = "additive",
lower_frequency_cutoff = 0.05,
upper_frequency_cutoff = 0.98,
correction = "bonferroni",
exponentiate = TRUE
) %>% filter(p.adjusted<0.05)
test_0331.csv
test_phenotype_0331.csv

[xiw588]

Hi, any ideas ?

[xiw588]

• Warnings:
• Using formula(x) is deprecated when x is a character vector of length > 1.
Consider formula(paste(x, collapse = " ")) instead.; occured 135 times, including variables: A01:01, A01:02, A02:01, A02:02, A*02:05, ...

[xiw588]

Hi, I am checking back to see if you have any ideas of how to fix it. Thanks

[mcjmigdal]

It is hard to pin down why these warning pops up, but the result is produced so I think you can just ignore them.

library(midasHLA)
library(survival)
library(dplyr)
test_0331 <- read.csv('test_0331.csv')
phenotype_data <- read.csv('test_phenotype_0331.csv')
hla_midas_alleles<-as.data.frame(test_0331)
HLA <- prepareMiDAS(
hla_calls = hla_midas_alleles,
colData = phenotype_data,
experiment = c('hla_alleles','hla_supertypes','hla_het','hla_divergence')
)
surv_GIA<-coxph(Surv(age_diag, diagnosis_status) ~ term+Predicted_Pop+sex+smoking, data =HLA)
surv_GIA_allele_uncond <- runMiDAS(
object = surv_GIA,
experiment = "hla_alleles",
inheritance_model = "additive",
lower_frequency_cutoff = 0.05,
upper_frequency_cutoff = 0.98,
correction = "bonferroni",
exponentiate = TRUE
)

Warning messages:
1: Alleles B*13:02, B*13:02 could not be found in the alignment coercing to NA 
2: Alleles B*35:01, B*35:01 could not be found in the alignment coercing to NA 
3: Alleles B*35:02, B*35:02 could not be found in the alignment coercing to NA 
4: Alleles B*18:01, B*18:01 could not be found in the alignment coercing to NA 
Warning messages:
1: 'experiments' dropped; see 'drops()' 
2: While evaluating the model problems occured:

• Warnings:
• Using formula(x) is deprecated when x is a character vector of length > 1.
  Consider formula(paste(x, collapse = " ")) instead.; occured 35 times, including variables: A*01:01, A*02:01, A*03:01, A*24:02, B*07:02, ... 

surv_GIA_allele_uncond
# A tibble: 35 × 10
   allele     p.value p.adjusted estimate std.error conf.low conf.high statistic
   <chr>        <dbl>      <dbl>    <dbl>     <dbl>    <dbl>     <dbl>     <dbl>
 1 DRB1*03:01  0.0312          1  20.5        1.40   1.31      321.         2.15
 2 A*01:01     0.0414          1  27.2        1.62   1.14      650.         2.04
 3 DQA1*01:03  0.0473          1  30.1        1.72   1.04      870.         1.98
 4 DQB1*06:03  0.0473          1  30.1        1.72   1.04      870.         1.98
 5 DRB1*13:01  0.0473          1  30.1        1.72   1.04      870.         1.98
 6 DQA1*05:01  0.0473          1  30.1        1.72   1.04      870.         1.98
 7 DQA1*02:01  0.0479          1   0.0571     1.45   0.00335     0.973     -1.98
 8 DRB1*07:01  0.0479          1   0.0571     1.45   0.00335     0.973     -1.98
 9 C*07:01     0.0602          1   5.57       0.913  0.929      33.3        1.88
10 C*04:01     0.0750          1   0.161      1.03   0.0215      1.20      -1.78
# ℹ 25 more rows
# ℹ 2 more variables: Ntotal <dbl>, Ntotal.percent <formttbl>
# ℹ Use `print(n = ...)` to see more rows

[xiw588]

Thanks! A follow up questions about the estimate of time to run the analysis. I was planning to run a multivariable (adjusting for sex, age, smoking and 10PCs) logistic regression in 32 alleles and corresponding amino acids in 30000 samples (roughly half half case and controls). It's been running for 5mins and froze once.... I wonder if something is wrong or should I just not to run this locally...Thank!

[mcjmigdal]

Oh that is an interesting question, sadly I do not have immediate answer to. Did you monitor your RAM usage during these 5 minutes?

[xiw588]

Haha, do you know how to monitor RAS usage for R studio locally?

[mcjmigdal]

If you are on Linux for you can use htop or top command, among others it will show your RAM usage. I would only check if your RAM is fully consumed or not. If yes then a bigger machine could solve your issue.

…

On Thu, Apr 17, 2025, 16:19 xiw588 ***@***.***> wrote: Haha, do you know how to monitor RAS usage for R studio locally? — Reply to this email directly, view it on GitHub <#48 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFOHDTEC5SZC4U63C5GJDWT2Z6Z75AVCNFSM6AAAAABZY7Z646VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMJTGEYDKMRXGU> . You are receiving this because you commented.Message ID: ***@***.***> *xiw588* left a comment (Genentech/midasHLA#48) <#48 (comment)> Haha, do you know how to monitor RAS usage for R studio locally? — Reply to this email directly, view it on GitHub <#48 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AFOHDTEC5SZC4U63C5GJDWT2Z6Z75AVCNFSM6AAAAABZY7Z646VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMJTGEYDKMRXGU> . You are receiving this because you commented.Message ID: ***@***.***>