Error in get_snps_frquncies_coverage function during hapcorrect module

[pabloangulo7]

Hi,

Thanks for the tool. I'm trying to run the hapcorrect module for my sample by using as input the bam file, reference genome fasta and the phased SNP vcf file in tumor only-mode, with this command:

python ${wakhan} hapcorrect --threads ${threads} --reference ${ref_genome_fasta} --target-bam ${bam_file} --tumor-phased-vcf ${snp_phased_vcf} --genome-name ${sample} --out-dir-plots ${HAP_correction}

It seems that it works well for all chromosomes except for chr14 and chr15 where I encounter this error. Which could be the reason?

[2025-11-16 23:56:53] INFO: Starting Wakhan 0.2.0
[2025-11-16 23:56:53] INFO: Cmd: /home/pangulo/Software/Wakhan/wakhan.py hapcorrect --threads 20 --reference /Test/GRCh38.primary_assembly.genome.fa --target-bam /Test/Sample.bam --tumor-phased-vcf /Test/Sample_snp_phased.vcf.gz --genome-name Sample --out-dir-plots /projects/CGS_shared/pangulo/BC/DNA/analysis/SNP/Hap_correction2/
[2025-11-16 23:56:53] INFO: Python version: 3.8.0 | packaged by conda-forge | (default, Nov 22 2019, 19:11:38) 
[GCC 7.3.0]
[2025-11-16 23:56:55] INFO: Starting hapcorrect() module...
[2025-11-16 23:56:55] INFO: Parsing reads from /Test/Sample.bam
[2025-11-17 00:07:56] INFO: Parsed 40225099 segments
[2025-11-17 00:07:56] INFO: Computing coverage histogram
[2025-11-17 00:14:20] INFO: Writing tumor coverage for bins
[2025-11-17 00:14:20] INFO: Parsing phaseblocks information
[2025-11-17 00:14:20] INFO: bcftools -> Query for phasesets and GT, DP, VAF feilds by creating a CSV file
[2025-11-17 00:14:36] INFO: Computing coverage for bins
[2025-11-17 00:14:37] INFO: bcftools -> Query for het SNPs and creating a /Test/Sample_snp_phased.vcf_het_snps.csv CSV file
[2025-11-17 00:14:44] INFO: SNPs frequency -> CSV to dataframe conversion for heterozygous SNPs
[2025-11-17 00:14:49] INFO: SNPs frequency -> Computing SNPs frequency from tumor BAM
[2025-11-17 00:25:16] INFO: SNPs frequency -> Computing ACGTs frequencies for heterozygous SNPs
[2025-11-17 00:26:52] INFO: bcftools -> Query for phasesets and GT, DP, VAF feilds by creating a CSV file
/home/pangulo/Software/Wakhan/src/hapcorrect/src/utils.py:62: DtypeWarning: Columns (5) have mixed types. Specify dtype option on import or set low_memory=False.
  dataframe = pd.read_csv(path, sep=sept, names=names)
[2025-11-17 00:27:15] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr1
[2025-11-17 00:28:33] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr2
[2025-11-17 00:29:12] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr3
[2025-11-17 00:29:38] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr4
[2025-11-17 00:30:44] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr5
[2025-11-17 00:31:30] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr6
[2025-11-17 00:31:53] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr7
[2025-11-17 00:32:37] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr8
[2025-11-17 00:33:01] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr9
[2025-11-17 00:33:20] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr10
[2025-11-17 00:33:37] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr11
[2025-11-17 00:34:03] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr12
[2025-11-17 00:34:32] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr13
[2025-11-17 00:34:41] INFO: Loading coverage (bins) and coverage (phaseblocks) datasets for chr14
Traceback (most recent call last):
  File "/home/pangulo/Software/Wakhan/wakhan.py", line 28, in <module>
    main()
  File "/home/pangulo/Software/Wakhan/wakhan.py", line 24, in main
    sys.exit(main())
  File "/home/pangulo/Software/Wakhan/src/main.py", line 353, in main
    main_process(args)  # hapcorrect()
  File "/home/pangulo/Software/Wakhan/src/hapcorrect/src/main_hapcorrect.py", line 234, in main_process
    get_snps_frquncies_coverage(df_snps_in_csv_loh, chrom, ref_start_values, args.bin_size_snps,
  File "/home/pangulo/Software/Wakhan/src/hapcorrect/src/process_vcf.py", line 155, in get_snps_frquncies_coverage
    snps_df_vaf = [eval(i) for i in snps_df.vaf.str.split(',').str[0].values.tolist()]
  File "/home/pangulo/Software/Wakhan/src/hapcorrect/src/process_vcf.py", line 155, in <listcomp>
    snps_df_vaf = [eval(i) for i in snps_df.vaf.str.split(',').str[0].values.tolist()]
TypeError: eval() arg 1 must be a string, bytes or code object 

Thank you very much in advanced,
Pablo

[pabloangulo7]

I think I identified the error. Apparently in the code, after this command:
df_snps_in_csv_loh = csv_df_chromosomes_sorter(output_phasesets_file_path, ['chr', 'pos', 'qual', 'gt', 'dp', 'vaf'])
The vaf column has mixed types between str and floats. This leads to an error in the command:
snps_df_vaf = [eval(i) for i in snps_df.vaf.str.split(',').str[0].values.tolist()]

So, I guess that in order to solve the issue I would need to convert the whole column of vaf to str

[tahashmi]

Hi @pabloangulo7 , Thanks for looking into it, I think that column is already being converted to str, maybe some entry is malfunctioned in VCF of these chroms?

[pabloangulo7]

Hi @tahashmi, thank you. I checked the vcf file and I didn't notice anything strange. The only difference to the recommended pipeline is that I first do the variant calling with Clair3 and then I phase the SNPs using whatshap, so maybe the input can have variations compared to longphase one but I'm not sure. In any case, I managed to run the hapcorrect module succesfully just by adding one line to the utils.py script in the csv_df_chromosomes_sorter function:

def csv_df_chromosomes_sorter(path, names, sept='\t'):
    dataframe = pd.read_csv(path, sep=sept, names=names)
    dataframe['chr'] = dataframe['chr'].astype(str)
    if 'vaf' in names:
        dataframe['vaf'] = dataframe['vaf'].astype(str) ##This line to convert the column to str

Based on the result, I would have one doubt (more conceptual) so maybe it's better that I open a new issue.

Thanks,
Pablo

[mikolmogorov]

Thanks @pabloangulo7! In principle, there could be library / Python differences affecting type conversion. Do you mind sending a small sample file so that we could reproduce on our end?

@tahashmi have we integrated this yet?

[Hugbenny]

I think the problem comes from the AF FORMAT of clair3. The vaf are written with "." as separator, which seems to not be converted correctly inside panda :

#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  SAMPLE
chr1    249     .       C       A       15.48   PASS    F       GT:GQ:DP:AD:AF:PS       0|1:15:5:4,1:0.2:249
chr1    767     .       A       G       11.79   PASS    F       GT:GQ:DP:AD:AF:PS       0|1:11:5:4,1:0.2:249
chr1    882     .       C       G       5.08    PASS    F       GT:GQ:DP:AD:AF:PS       0|1:5:5:4,1:0.2:249

I bypassed this by replacing :

• in processing.py line 95 :

# haplotype_1_coverage = [eval(i) for i in df.vaf.str.split(',').str[0].values.tolist()]
haplotype_1_coverage = pd.to_numeric(df.vaf.str.split(',').str[0], errors='coerce').values.tolist()

• in process_vcf.py line 55 +98 + 132 :

# snps_df_vaf = [eval(i) for i in snps_df.vaf.str.split(',').str[0].values.tolist()]
snps_df_vaf = pd.to_numeric(snps_df.vaf.str.split(',').str[0], errors='coerce').values.tolist()

Also, i could suggest you to progressively replace panda by polars for the speed, it's way faster.

Very good job on this tool !

[mikolmogorov]

@Hugbenny thanks - are you running the most recent version and it still giving this error? If so, do you mind sending us an example input that breaks Wakhan? (just the vcf should be fine!). We haven't seen issues with Clair3 on our end previously..

[Hugbenny]

Sure, i can send you the vcf by mail, no patient data since it's from simulated bam. I ran the 0.4.2 version installed via bioconda.