Bio tumor fraction estimation

Version Compatibility

Reference examples tested with: CNVkit 0.9+, ichorCNA 0.5+, pandas 2.2+

Before using code patterns, verify installed versions match. If versions differ:

• Python: pip show <package> then help(module.function) to check signatures
• R: packageVersion('<pkg>') then ?function_name to verify parameters

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

Tumor Fraction Estimation

"Estimate tumor fraction from my cfDNA data" → Calculate the proportion of tumor-derived DNA in a liquid biopsy sample using copy number aberrations from shallow whole-genome sequencing.

• R: ichorCNA for tumor fraction and CNA estimation from sWGS

Estimate ctDNA tumor fraction from shallow whole-genome sequencing.

ichorCNA Overview

ichorCNA (GavinHaLab fork, v0.5.1+) detects copy number alterations and estimates tumor fraction from sWGS (0.1-1x coverage).

Sensitivity: 97-100% detection at >= 3% tumor fraction (2024 validation)

Input Requirements

Requirement	Specification
Data type	sWGS (NOT targeted panel)
Coverage	0.1-1x (0.5x recommended)
Input	BAM files
Output	Tumor fraction, ploidy, CNA segments

Running ichorCNA

library(ichorCNA)

# Step 1: Generate read counts in bins
# Run from command line or use HMMcopy
# readCounter --window 1000000 --quality 20 sample.bam > sample.wig

# Step 2: Run ichorCNA
runIchorCNA(
    WIG = 'sample.wig',
    gcWig = 'gc_hg38_1mb.wig',
    mapWig = 'mappability_hg38_1mb.wig',
    normalPanel = 'pon_median_1mb.rds',
    centromere = 'centromeres_hg38.txt',
    outDir = 'ichor_results/',
    id = 'sample_id',

    # Tumor fraction estimation parameters
    normal = c(0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99),
    ploidy = c(2, 3),
    maxCN = 5,

    # Subclonality
    estimateScPrevalence = TRUE,
    scStates = c(1, 3),

    # Segmentation
    txnE = 0.9999,
    txnStrength = 10000,

    # Chromosomes
    chrs = paste0('chr', c(1:22, 'X'))
)

Batch Processing

Goal: Run ichorCNA tumor fraction estimation on a cohort of sWGS samples in parallel, collecting results and handling failures gracefully.

Approach: Apply the ichorCNA pipeline to each sample's WIG file using mclapply for parallelization, wrapping each call in tryCatch to report per-sample success or failure.

library(ichorCNA)
library(parallel)

process_sample <- function(wig_file, params) {
    sample_id <- basename(wig_file)
    sample_id <- gsub('.wig$', '', sample_id)

    tryCatch({
        runIchorCNA(
            WIG = wig_file,
            gcWig = params$gcWig,
            mapWig = params$mapWig,
            normalPanel = params$normalPanel,
            centromere = params$centromere,
            outDir = params$outDir,
            id = sample_id,
            normal = c(0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99),
            ploidy = c(2, 3),
            maxCN = 5
        )
        return(list(sample = sample_id, status = 'success'))
    }, error = function(e) {
        return(list(sample = sample_id, status = 'failed', error = e$message))
    })
}

# Run in parallel
wig_files <- list.files('wig/', pattern = '.wig$', full.names = TRUE)
params <- list(
    gcWig = 'gc_hg38_1mb.wig',
    mapWig = 'mappability_hg38_1mb.wig',
    normalPanel = 'pon_median_1mb.rds',
    centromere = 'centromeres_hg38.txt',
    outDir = 'ichor_results/'
)

results <- mclapply(wig_files, process_sample, params = params, mc.cores = 4)

Parsing Results

parse_ichor_results <- function(results_dir) {
    # Find results files
    param_files <- list.files(results_dir, pattern = '.params.txt$',
                              full.names = TRUE, recursive = TRUE)

    results <- data.frame()

    for (f in param_files) {
        params <- read.table(f, header = TRUE, sep = '\t', stringsAsFactors = FALSE)
        sample_id <- gsub('.params.txt$', '', basename(f))

        results <- rbind(results, data.frame(
            sample = sample_id,
            tumor_fraction = 1 - params$n[1],  # n is normal fraction
            ploidy = params$phi[1],
            log_likelihood = params$loglik[1]
        ))
    }

    return(results)
}

# Parse all results
tf_results <- parse_ichor_results('ichor_results/')
print(tf_results)

Python Wrapper

import subprocess
import pandas as pd
from pathlib import Path


def run_ichorcna(wig_file, output_dir, gc_wig, map_wig, normal_panel, centromere):
    '''Run ichorCNA from Python.'''
    sample_id = Path(wig_file).stem

    cmd = f'''
    Rscript -e "
    library(ichorCNA)
    runIchorCNA(
        WIG = '{wig_file}',
        gcWig = '{gc_wig}',
        mapWig = '{map_wig}',
        normalPanel = '{normal_panel}',
        centromere = '{centromere}',
        outDir = '{output_dir}',
        id = '{sample_id}',
        normal = c(0.5, 0.6, 0.7, 0.8, 0.9, 0.95, 0.99),
        ploidy = c(2, 3),
        maxCN = 5
    )
    "
    '''

    subprocess.run(cmd, shell=True, check=True)


def parse_tumor_fraction(params_file):
    '''Parse tumor fraction from ichorCNA output.'''
    df = pd.read_csv(params_file, sep='\t')
    return {
        'tumor_fraction': 1 - df['n'].iloc[0],
        'ploidy': df['phi'].iloc[0],
        'log_likelihood': df['loglik'].iloc[0]
    }

Interpretation

Tumor Fraction	Interpretation
>= 10%	High ctDNA, reliable detection
3-10%	Moderate ctDNA, detectable
< 3%	Low ctDNA, at detection limit
0%	No detectable ctDNA or below LOD

Related Skills

• cfdna-preprocessing - Preprocess BAMs before ichorCNA
• fragment-analysis - Complementary fragmentomics analysis
• ctdna-mutation-detection - Mutation detection from panel data
• copy-number/cnvkit-analysis - CNV concepts