Error under Regression Models

Shann Jiang

unread,

Jun 17, 2016, 11:52:07 PM6/17/16

to PRSice

I ran a simple PRSice and encountered the error info as follows:

Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, :
NA/NaN/Inf in 'x'
Calls: summary ... with.default -> eval -> eval -> glm -> eval -> eval -> glm.fit
Execution halted

Any one know what's going on?

My command:

R -q --file=./PRSice_v1.25.R --args \
> base NLandADcarriers-v2.assoc \
> target NLandADnoncarriers-v2 \
> slower 0 \
> supper 0.5 \
> sinc 0.01 \
> covary F \
> clump.snps F \
> plink ./plink_1.9_linux_160914 \
> figname carriervsnoncarrier
> library(batch)
> start.time <- proc.time()[3]
> options(echo = FALSE)
#################################
#
#
#
#
# PRSice: Polygenic Risk Score software
#
# Jack Euesden, Cathryn M. Lewis, Paul F. O'Reilly 2014
#
#
# If you use PRSice in published work, please cite:
#
# "PRSice: Polygenic Risk Score software"
# Euesden, Lewis, O'Reilly, Bioinformatics (2015) 31 (9):1466-1468
#
#
#
#
#################################
#################################
#
# Read in Command Line Arguments & interpret
#
#################################
$base
[1] "NLandADcarriers-v2.assoc"

$target
[1] "NLandADnoncarriers-v2"

$slower
[1] 0

$supper
[1] 0.5

$sinc
[1] 0.01

$covary
[1] "F"

$clump.snps
[1] "F"

$plink
[1] "./plink_1.9_linux_160914"

$figname
[1] "carriervsnoncarrier"

#################################
#
# Check options match
#
#################################
[1] "Using current directory as working directory"
#################################
#
# Check base input format
#
#################################
[1] "WARNING: No SE column in base data"
#################################
#
#   Remove Ambiguous SNPs
#
#################################
#################################
#
#   Deal with strand flips if target is in genotype format and produce input files for polygenic scoring
#
#################################
#################################
#
#   Polygenic scoring!
#
#################################
#################################
#
#   Regression Models
#
#################################
Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, :
NA/NaN/Inf in 'x'
Calls: summary ... with.default -> eval -> eval -> glm -> eval -> eval -> glm.fit
Execution halted

Tulsi Patel

unread,

Jun 27, 2016, 10:10:40 AM6/27/16

to PRSice

I have been having the same problem - with a slightly different error message now.

#################################

#

# Regression Models

#

#################################

Error in family$linkfun(mustart) :

Argument mu must be a nonempty numeric vector

Calls: glm -> eval -> eval -> glm.fit -> <Anonymous> -> .Call

Execution halted

This is the command I used:

R -q --file=./PRSice_v1.25.R --args base data.assoc target data2 slower 0 supper 0.5 sinc 0.01 plink ./plink_1.9_mac_160914 fig name output

Anyone who has any idea how to fix this, it would be very very much appreciated :)

Joni Coleman

unread,

Aug 25, 2016, 9:15:43 AM8/25/16

to PRSice

Although this thread has been cold for a while, I will offer a little insight in case others are encountering this problem.

Both of these problems are likely to be related to the phenotype file, and particularly to missing data or multicollinearity in the phenotypes / covariates.

If either of the posters is still having this problem, I may be help if you could post an example of your phenotype file.

Sultan Chaudhury

unread,

Dec 2, 2016, 10:16:14 AM12/2/16

to PRSice

Hi,

I am also having this issue with the exact same message as Tulsi.
I have not included a phenotype nor covariates file so I am unsure of where the source of the issue is.

Thanks,
Sultan

Shing Wan Choi

unread,

Jan 1, 2017, 9:30:11 AM1/1/17

to PRSice

If you have not included a phenotype nor covariates file, then it is very likely that your fam file contains only 1 phenotype (e.g. all control). In that case, we cannot perform the GLM analysis, therefore the error message.

S Sam

unread,

Feb 1, 2017, 12:30:08 PM2/1/17

to PRSice

Hi Joni and Shing Wang and others,

I have been facing the exact same issue.

#################################

#

# Regression Models

#

#################################

Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, :

NA/NaN/Inf in 'x'

Calls: summary ... with.default -> eval -> eval -> glm -> eval -> eval -> glm.fit

Execution halted

- I dont have any missing phenotype data. Although I have not supplied any phenotype file. A snap shot of my .fam file is as follows :

25 A001 0 0 2 1

27 A004 0 0 2 2

The phenotype in the 6th column of the .fam file is coded as 1/2.

I have not supplied any covariate file and have used the command "covary F ".

Any further advice to trouble shoot this issue ?

Many thanks

Jonathan Coleman

unread,

Feb 1, 2017, 12:35:55 PM2/1/17

to PRSice

Hi S Sam,

Can you post the full output file (as in the first poster)?

Joni

S Sam

unread,

Feb 1, 2017, 12:40:24 PM2/1/17

to PRSice

Hi Joni,

Thank you for your response. Please see below.

My script:

R -q --file=./PRSice_v1.25.R --args \

plink ./plink_1.9_linux_160914 \

base /mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/output/imputed_dataset/Sept2016_Re_QCd/acr30NR_30R/acr30R_NR_imputed_gwas_maf_1_perc.assoc \

target /mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/data/clean/IMPUTED_DATA/RE_QC_Sept_2016/FINAL_SEPT2016_DATASET/acr_30R_NR_only_imputed_updated_pheno \

slower 0 \

supper 0.5 \

sinc 0.01 \

covary F \

figname ACR30R_30NR

Output:

#

#################################

#

# Read in Command Line Arguments & interpret

#

#################################

$plink

[1] "./plink_1.9_mac_160914"

$base

[1] "/mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/output/imputed_dataset/Sept2016_Re_QCd/acr30NR_30R/acr30R_NR_imputed_gwas_maf_1_perc.assoc"

$target

[1] "/mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/data/clean/IMPUTED_DATA/RE_QC_Sept_2016/FINAL_SEPT2016_DATASET/acr_30R_NR_only_imputed_updated_pheno"

$slower

[1] 0

$supper

[1] 0.5

$sinc

[1] 0.01

$covary

[1] "F"

$figname

[1] "ACR30R_30NR"

#################################

#

# Check options match

#

#################################

[1] "Using current directory as working directory"

#################################

#

# Check base input format

#

#################################

[1] "WARNING: No SE column in base data"

#################################

#

# Remove Ambiguous SNPs

#

#################################

sh: ./plink_1.9_mac_160914: No such file or directory

#################################

#

# Clump

#

#################################

awk: cmd. line:1: fatal: cannot open file `non_synonymous_snps_only.bim' for reading (No such file or directory)

sh: ./plink_1.9_mac_160914: No such file or directory

tail: cannot open `cleaned_base.clumped' for reading: No such file or directory

#################################

#

# Deal with strand flips if target is in genotype format and produce input files for polygenic scoring

#

#################################

awk: cmd. line:1: fatal: cannot open file `non_synonymous_snps_only.bim' for reading (No such file or directory)

Error in read.table("Complete_Allele_List.txt", head = F) :

no lines available in input

Execution halted

[mbchkssb@incline05 PRSice_v1.25]$ module load apps/gcc/R/3.3.0

[mbchkssb@incline05 PRSice_v1.25]$ chmod uga+wrx ./plink_1.9_linux_160914

[mbchkssb@incline05 PRSice_v1.25]$ chmod uga+wrx ./PRSice_v1.25.R

[mbchkssb@incline05 PRSice_v1.25]$ R -q --file=./PRSice_v1.25.R --args \

> plink ./plink_1.9_linux_160914 \

> base /mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/output/imputed_dataset/Sept2016_Re_QCd/acr30NR_30R/acr30R_NR_imputed_gwas_maf_1_perc.assoc \

> target /mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/data/clean/IMPUTED_DATA/RE_QC_Sept_2016/FINAL_SEPT2016_DATASET/acr_30R_NR_only_imputed_updated_pheno \

> slower 0 \

> supper 0.5 \

> sinc 0.01 \

> covary F \

> figname ACR30R_30NR

$plink

[1] "./plink_1.9_linux_160914"

$base

[1] "/mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/output/imputed_dataset/Sept2016_Re_QCd/acr30NR_30R/acr30R_NR_imputed_gwas_maf_1_perc.assoc"

$target

[1] "/mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/data/clean/IMPUTED_DATA/RE_QC_Sept_2016/FINAL_SEPT2016_DATASET/acr_30R_NR_only_imputed_updated_pheno"

$slower

[1] 0

$supper

[1] 0.5

$sinc

[1] 0.01

$covary

[1] "F"

$figname

[1] "ACR30R_30NR"

#################################

#

# Check options match

#

#################################

[1] "Using current directory as working directory"

#################################

#

# Check base input format

#

#################################

[1] "WARNING: No SE column in base data"

#################################

#

# Remove Ambiguous SNPs

#

#################################

#

# Clump

#

#################################

#

# Deal with strand flips if target is in genotype format and produce input files for polygenic scoring

#

#################################

#

# Polygenic scoring!

#

#################################

S Sam

unread,

Feb 1, 2017, 12:45:01 PM2/1/17

to PRSice

Hi Joni,

sorry, what I posted earlier was slightly wrong (I included an earlier error, which I rectified).Please see below for correct output.

My script:

R -q --file=./PRSice_v1.25.R --args \

plink ./plink_1.9_linux_160914 \

base /mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/output/imputed_dataset/Sept2016_Re_QCd/acr30NR_30R/acr30R_NR_imputed_gwas_maf_1_perc.assoc \

target /mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/data/clean/IMPUTED_DATA/RE_QC_Sept_2016/FINAL_SEPT2016_DATASET/acr_30R_NR_only_imputed_updated_pheno \

slower 0 \

supper 0.5 \

sinc 0.01 \

covary F \

figname ACR30R_30NR

PRSice: Polygenic Risk Score software

#

# Jack Euesden, Cathryn M. Lewis, Paul F. O'Reilly 2014

#

# If you use PRSice in published work, please cite:

#

# "PRSice: Polygenic Risk Score software"

# Euesden, Lewis, O'Reilly, Bioinformatics (2015) 31 (9):1466-1468

#

#################################

#

# Read in Command Line Arguments & interpret

#

#################################

$plink

[1] "./plink_1.9_linux_160914"

$base

[1] "/mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/output/imputed_dataset/Sept2016_Re_QCd/acr30NR_30R/acr30R_NR_imputed_gwas_maf_1_perc.assoc"

$target

[1] "/mnt/jw01-aruk-home01/projects/jia_mtx_gwas_2016/common_files/data/clean/IMPUTED_DATA/RE_QC_Sept_2016/FINAL_SEPT2016_DATASET/acr_30R_NR_only_imputed_updated_pheno"

$slower

[1] 0

$supper

[1] 0.5

$sinc

[1] 0.01

$covary

[1] "F"

$figname

[1] "ACR30R_30NR"

#################################

#

# Check options match

#

#################################

[1] "Using current directory as working directory"

#################################

#

# Check base input format

#

#################################

[1] "WARNING: No SE column in base data"

#################################

#

# Remove Ambiguous SNPs

#

#################################

#

# Clump

#

#################################

#

# Deal with strand flips if target is in genotype format and produce input files for polygenic scoring

#

#################################

#

# Polygenic scoring!

#

#################################

Jonathan Coleman

unread,

Feb 1, 2017, 12:50:27 PM2/1/17

to PRSice

Ok - this could be caused by missing phenotypes in your .fam file. Please check that. If you don't have missing phenotypes, please re-run including the command "debug.mode T" (without quotes) and post the output.debug.modedebug.mode

S Sam

unread,

Feb 1, 2017, 1:08:25 PM2/1/17

to PRSice

I have checked .fam file and none are mmissing phenotypes. Here is the output produced in the debug mode.

Thank you.

$debug.mode

[1] "T"

$figname

[1] "ACR30R_30NR"

#################################

#

# Check options match

#

#################################

[1] "Using current directory as working directory"

#################################

#

# Check base input format

#

#################################

[1] "WARNING: No SE column in base data"

#################################

#

# Remove Ambiguous SNPs

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to non_synonymous_snps_only.log.

Note: --noweb has no effect since no web check is implemented yet.

64496 MB RAM detected; reserving 32248 MB for main workspace.

7294701 variants loaded from .bim file.

971 people (315 males, 656 females) loaded from .fam.

971 phenotype values loaded from .fam.

--exclude: 6224150 variants remaining.

Warning: At least 57407 duplicate IDs in --exclude file.

Using 1 thread (no multithreaded calculations invoked).

Calculating allele frequencies... done.

6224150 variants and 971 people pass filters and QC.

Among remaining phenotypes, 267 are cases and 704 are controls.

--make-bed to non_synonymous_snps_only.bed + non_synonymous_snps_only.bim +

non_synonymous_snps_only.fam ... done.

#################################

#

# Clump

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to cleaned_base.log.

Note: --noweb has no effect since no web check is implemented yet.

64496 MB RAM detected; reserving 32248 MB for main workspace.

6224150 variants loaded from .bim file.

971 people (315 males, 656 females) loaded from .fam.

971 phenotype values loaded from .fam.

Using 1 thread (no multithreaded calculations invoked).

Calculating allele frequencies... done.

6224150 variants and 971 people pass filters and QC.

Among remaining phenotypes, 267 are cases and 704 are controls.

--clump: 283187 clumps formed from 5990087 top variants.

Results written to cleaned_base.clumped .

#################################

#

# Deal with strand flips if target is in genotype format and produce input files for polygenic scoring

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to flipped_target.log.

Note: --noweb has no effect since no web check is implemented yet.

64496 MB RAM detected; reserving 32248 MB for main workspace.

6224150 variants loaded from .bim file.

971 people (315 males, 656 females) loaded from .fam.

971 phenotype values loaded from .fam.

--flip: 0 SNPs flipped.

Using 1 thread (no multithreaded calculations invoked).

Calculating allele frequencies... done.

6224150 variants and 971 people pass filters and QC.

Among remaining phenotypes, 267 are cases and 704 are controls.

--make-bed to flipped_target.bed + flipped_target.bim + flipped_target.fam ...

done.

#################################

#

# Polygenic scoring!

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to PROFILES.log.

Note: --noweb has no effect since no web check is implemented yet.

64496 MB RAM detected; reserving 32248 MB for main workspace.

6224150 variants loaded from .bim file.

971 people (315 males, 656 females) loaded from .fam.

971 phenotype values loaded from .fam.

Using 1 thread (no multithreaded calculations invoked).

Calculating allele frequencies... done.

6224150 variants and 971 people pass filters and QC.

Among remaining phenotypes, 267 are cases and 704 are controls.

--score: 50 ranges processed (1 empty range skipped).

Results written to PROFILES.*.profile.

Nikman Adli Nor Hashim

unread,

Feb 19, 2017, 7:15:27 PM2/19/17

to PRSice

Hi Shann,

I am not the expert one, but I would like to help on this issue.

Can you do these and print them out for me please:

1. head .fam file

2. head .assoc file

3. In both fam and .assoc files, check your missing values again (in excel or wherever), how many are they? make sure all of them are denoted as 'NA' only (I had some missing values denoted as 'blank' before)

I reckon there shouldnt be any problem with .assoc file since the base-PRS was calculated. However when PRSice tried to regress it with your target data,it just couldnt. Hence, the focus would be on the target phenotype values.

Best wishes,

Nikman

Hui Wang

unread,

Feb 22, 2017, 9:58:44 AM2/22/17

to PRSice

Hi Shann and others,

I'm now confronted with this same issue now. Any explanations for this error and solutions?:)

Jonathan Coleman

unread,

Feb 22, 2017, 10:24:09 AM2/22/17

to PRSice

Please see my previous answer and Nikman's above:

Check your data for missingness
Check your data for multicollinearity (variables that predict other variables)

If there is no missingness and no multicollinearity, please provide the log output so we can assist.

Thanks,

Joni

Hui Wang

unread,

Feb 22, 2017, 11:43:06 AM2/22/17

to PRSice

Hi Jonathan,

I've checked, Phenotype targeted is coded as 1/2 in fam file, without missing values. And I don't apply any external phenotype data or covariates.

What confuses me more is when I use another base data instead of "converge" ,this error didn't occur.

As for debug,mode log, please see below,

[root@promote converge-MR_allSNP]# R --file=PRS.R -q --args plink /converge-MR_allSNP/plink_1.9_linux_160914 base /converge-MR_allSNP/converge target /converge-MR_allSNP/MRsample_all_SNP slower 0 sinc 0.01 supper 0.5 covary F debug.mode T

$plink

[1] "/converge-MR_allSNP/plink_1.9_linux_160914"

$base

[1] "/converge-MR_allSNP/converge"

$target

[1] "/converge-MR_allSNP/MRsample_all_SNP"

$slower

[1] 0

$sinc

[1] 0.01

$supper

[1] 0.5

$covary

[1] "F"

$debug.mode

[1] "T"

#################################

#

# Check options match

#

#################################

[1] "Using current directory as working directory"

#################################

#

# Check base input format

#

#################################

#

# Remove Ambiguous SNPs

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to non_synonymous_snps_only.log.

Note: --noweb has no effect since no web check is implemented yet.

976 MB RAM detected; reserving 488 MB for main workspace.

10555 variants loaded from .bim file.

90 people (47 males, 43 females) loaded from .fam.

90 phenotype values loaded from .fam.

--exclude: 10555 variants remaining.

Using 1 thread.

Calculating allele frequencies... done.

Warning: 300 het. haploid genotypes present (see non_synonymous_snps_only.hh ).

Total genotyping rate is 0.980723.

10555 variants and 90 people pass filters and QC.

Among remaining phenotypes, 54 are cases and 36 are controls.

--make-bed to non_synonymous_snps_only.bed + non_synonymous_snps_only.bim +

non_synonymous_snps_only.fam ... done.

#################################

#

# Clump

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to cleaned_base.log.

Note: --noweb has no effect since no web check is implemented yet.

976 MB RAM detected; reserving 488 MB for main workspace.

10555 variants loaded from .bim file.

90 people (47 males, 43 females) loaded from .fam.

90 phenotype values loaded from .fam.

Using 1 thread.

Calculating allele frequencies... done.

Warning: 300 het. haploid genotypes present (see cleaned_base.hh ).

Total genotyping rate is 0.980723.

10555 variants and 90 people pass filters and QC.

Among remaining phenotypes, 54 are cases and 36 are controls.

--clump: 2437 clumps formed from 8675 top variants.

Results written to cleaned_base.clumped .

#################################

#

# Deal with strand flips if target is in genotype format and produce input files for polygenic scoring

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to flipped_target.log.

Note: --noweb has no effect since no web check is implemented yet.

976 MB RAM detected; reserving 488 MB for main workspace.

10555 variants loaded from .bim file.

90 people (47 males, 43 females) loaded from .fam.

90 phenotype values loaded from .fam.

--flip: 2 SNPs flipped.

Using 1 thread.

Calculating allele frequencies... done.

Warning: 300 het. haploid genotypes present (see flipped_target.hh ).

Total genotyping rate is 0.980723.

10555 variants and 90 people pass filters and QC.

Among remaining phenotypes, 54 are cases and 36 are controls.

--make-bed to flipped_target.bed + flipped_target.bim + flipped_target.fam ...

done.

#################################

#

# Polygenic scoring!

#

#################################

PLINK v1.90b2i 64-bit (8 Sep 2014) https://www.cog-genomics.org/plink2

Logging to PROFILES.log.

Note: --noweb has no effect since no web check is implemented yet.

976 MB RAM detected; reserving 488 MB for main workspace.

10555 variants loaded from .bim file.

90 people (47 males, 43 females) loaded from .fam.

90 phenotype values loaded from .fam.

Using 1 thread.

Calculating allele frequencies... done.

Warning: 300 het. haploid genotypes present (see PROFILES.hh ).

Total genotyping rate is 0.980723.

10555 variants and 90 people pass filters and QC.

Among remaining phenotypes, 54 are cases and 36 are controls.

Warning: 2 lines skipped in --score file.

Warning: 2 lines skipped in --q-score-range data file.

--score: 50 ranges processed (1 empty range skipped).

Results written to PROFILES.*.profile.

#################################

#

# Regression Models

#

#################################

Regression Models: 10% Complete

Regression Models: 20% Complete

Regression Models: 30% Complete

Error in glm.fit(x = c(1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, :

NA/NaN/Inf in 'x'

Calls: summary ... with.default -> eval -> eval -> glm -> eval -> eval -> glm.fit

Execution halted

[root@promote converge-MR_allSNP]#

dajiangd...@gmail.com

unread,

Mar 6, 2017, 9:46:49 AM3/6/17

to PRSice

Hi everyone,

The error information indicates that x has -Inf values. x is score generated by PLINK --score command. As I can find in PLINK 1.0 doc. This --score command might use log(OR) as score for each SNP in training sample, which is subsequently used to generate Polygenic Score of each individual in replication sample.If so, one of the included ORs under certain P thresh must approach 0, then the log(OR) could be -Inf. But I check my base data, none of the ORs is 0. Anyone knows how exactly PLINK calculates score? May be we can figure out the source of this error by analyze the source code.

Hui

dajiangd...@gmail.com

unread,

Mar 9, 2017, 7:32:45 AM3/9/17

to PRSice

Hi everyone,

With the great help of Jack Euesden, author of PRSice, I've fixed this error. Please check your Base data, there must be some NA values in the OR col. Remove them and try again.

ankita lodhavia

unread,

Apr 26, 2020, 6:18:07 AM4/26/20

to PRSice

Hello

I am also facing the same issue :

Error in family$linkfun(mustart) :

Argument mu must be a nonempty numeric vector

My datafile contains some NA values but the columns that I use for my GLM model do not have any NA values.

Does this still affect the model?

Please help!

Sam Choi

unread,

Apr 27, 2020, 12:06:23 PM4/27/20

to PRSice

Could you please include the full log file for me to check what's going on? Thank you

Reply all

Reply to author

Forward