Fig. 5: Estimation of tumour fraction in plasma cell-free DNA. | Nature Communications

Fig. 5: Estimation of tumour fraction in plasma cell-free DNA.

From: MethylBERT enables read-level DNA methylation pattern identification and tumour deconvolution using a Transformer-based model

Fig. 5

All boxplots represent the median, the first and third quartiles, whereas the whiskers show the rest of the distribution. A Tumour purity estimation of simulated pseudo-bulk samples (n = 10) with a low percentage of tumour DNA. Distribution of absolute percentage error values in each method. B Comparison of ground-truth and estimated tumour purities in each method. For each comparison, Spearman’s correlation is given with a respective p-value. The error bars indicate a 95% confidence interval. The two-sided p-values were calculated with a null hypothesis that two samples do not have an ordinal correlation, as implemented in the Scipy package48. C, D Tumour cell fraction estimation results in ctDNA samples (C) from CRC patients and (D) PDAC patients (Supplementary Fig. 6 provides the sample size information). Both analyses include healthy donors as well. The median of estimated tumour purities in each stage is written at the top of the box plot. Two-sided Mann–Whitney–Wilxcoxon test p-values between each stage and healthy donors are denoted with stars. ‘*’, ‘**’ and ‘***’ mean p-value ≤ 0.05, ≤ 0.01, and ≤ 0.001, respectively.

Back to article page