NOISY SEGMENTATION in most samples

[WangKang-Leo]

Describe the issue I followed your suggestion, run the PureCN from scratch. The distribution of purity becomes better, but it is commented on by Noisy Segmentation in most of the samples. Could you please help to look into one sample. To Reproduce

Step1 reference files

Rscript $PURECN/IntervalFile.R --in-file /proj/sens2019581/nobackup/PureCN/reference_files/Twist_Exome_RefSeq_targets_hg38_100bp_padding.bed \ --fasta /sw/data/uppnex/ToolBox/hg38bundle/Homo_sapiens_assembly38.fasta \ --genome hg38 \ --mappability /proj/sens2019581/nobackup/PureCN/reference_files/GCA_000001405.15_GRCh38_no_alt_analysis_set_100.bw \ --out-file $OUT_REF/Twist_Exome_RefSeq_targets_hg38_intervals.txt

Step2 Coverage

Rscript $PURECN/Coverage.R --out-dir $OUT \ --bam /proj/sens2019581/nobackup/PureCN/data/normal_bam.list \ --intervals $OUT_REF/Twist_Exome_RefSeq_targets_hg38_intervals.txt \ --cores 64 --parallel Rscript $PURECN/Coverage.R --out-dir $OUT \ --bam /proj/sens2019581/nobackup/PureCN/data/tumor_bam.list \ --intervals $OUT_REF/Twist_Exome_RefSeq_targets_hg38_intervals.txt \ --cores 64 --parallel

NormalDB

Rscript $PURECN/NormalDB.R --out-dir $OUT_REF \ --coverage-files /proj/sens2019581/nobackup/PureCN/data/Normal/BEVPAC_normal_coverages.list \ --genome hg38 --assay Twist_Exome

PureCN run

Rscript $PURECN/PureCN.R --out $OUT/$Sample \ --tumor /proj/sens2019581/nobackup/PureCN/data/Tumor/$Sample".recal_coverage_loess.txt.gz" \ --sampleid $Sample \ --vcf /proj/sens2019581/nobackup/PureCN/data/$Sample".vcf" \ --fun-segmentation PSCBS \ --normaldb $OUT_REF/normalDB_Twist_Exome_hg38.rds \ --intervals $OUT_REF/Twist_Exome_RefSeq_targets_hg38_intervals.txt \ --snp-blacklist $OUT_REF/hg38-blacklist.v2.bed \ --genome hg38 \ --minpurity 0.05 --max-copy-number 8 \ --rds=rda \ --max-non-clonal 0.3 \ --model betabin \ --force --post-optimize --seed 123 \ --cores 32

Log file UE-2971-1106-0.log UE-2971-1106-0.csv

B-allele frequency plot

[lima1]

This looks pretty decent. You can just increase the --max-segments. For whole exome sequencing of tissue, GATK's segmentation is worth trying.

[lima1]

--fun-segmentation GATK (with gatk in path).

[WangKang-Leo]

Dear author

Thanks for your prompt reply.

Here I ran the same sample using GATK4 segmentation and set --max-segments as 350. I found it outputs same purity and ploidy, but no "NOISY SEGMENTATION" comment and fewer LOH event (segments). Does it look better than before? I can use them for production?

Thanks/Kang UE-2971-1106-0.log

[lima1]

Looks good!