Fig. 3

Multisite sampling in GBM does not affect the GMC but aids in the detection of PGMs. (A), Differences in gene mutation coverage in multisite sampling. (B), Difference in gene mutation coverage following unsupervised consensus clustering on the basis of the CBV and CE-T1W images (n = 29). (C), Correlation analysis between gene mutation coverage and the distance between the sites in two-site sampling (n = 56). (D), Correlation analysis between the gene mutation coverage and the radius of the minimum covering circle (n = 54). (E), Correlation analysis between the number of PGMs and the distance between the sites of two-site sampling (n = 56). (F), Correlation analysis between the number of PGMs and the radius of the minimum covering circle (n = 54). G, Schematic diagram of the “vegetation” hypothesis for detecting homogeneity and heterogeneity in an intraoperative sample; the soil (brown) refers to SGMs, and vegetation refers to PGMs (specifically, trees, bushes and grasses refer to PGMs with high, moderate and low frequencies, respectively). The amount of soil and vegetation represents the number of corresponding gene mutations. An uneven “vegetation” distribution means that the detection of PGM will be different for each sample. GMC, gene mutation coverage; UGM, unique gene mutation; PGM, private gene mutation; GBM, glioblastoma; GMH, gene mutation heterogeneity.