Figure 1.
Figure 1. Large aberrations and small mutations (SNVs and indels) shape the MPN mutational landscape. (A) Distribution of diagnosis within the MPN cohort (outer ring). Occurrence of the MPN driver mutation status within each diagnosis (inner ring). Number of patients with each diagnosis (center). Healthy controls are in gray. (B) Transcriptome sequencing was performed on granulocyte RNA from 113 MPN patients and 15 healthy controls. Independent workflows for fusion calling, variant calling (SNVs and indels), and differential splicing analysis (in SF3B1-mutated patients) were established. Mutations were called, filtered, and validated; those leading to protein changes were translated to amino acid sequences. Neoantigens were predicted on a personalized level, taking each patient’s MHCI haplotype into consideration. (C) A total of 123 MPN patient samples is depicted and sorted by diagnosis, MPN driver mutation status (JAK2, CALR, or MPL positive), and nondriver mutation frequency. Patient sample replicates (n = 10) sequenced across or with the same batch (A-E) are indicated in capital letters (A-J). Chromosomal aberrations panel: uniparental disomies (UPDs), deletions, and gains were called with Affymetrix SNP 6.0 arrays (supplemental Table 5). Fusion panel: fusions were private among patients and colored by rearrangement type. Two fusions were reported for patient P106A#A. For fusion calling, we combined the results of 3 fusion detection tools (deFuse, SOAPfuse, and TopHat-Fusion) to overcome algorithm-specific biases, a practice that is frequently suggested in fusion benchmarking studies.39,40 MPN driver mutation panel: MPN driver mutation status (JAK2, CALR, and MPL) was determined as described in supplemental Methods. SNVs and indels panel: genes are grouped by occurrence in pathways and by mutation frequency. Mutations are colored by mutation type, and a gradient for high and low SIFT score was applied (only for SNVs). Small rectangles with black frame enhance visibility for mutations with low SIFT score. Only those genes with ≥2 mutations across the cohort, with the exception of genes involved in the splicing machinery (SRSF2, SF3A1), are shown. Gene names in blue are part of the TruSight Myeloid Sequencing Panel. For validation, variants from 77 RNA/DNA pairs were compared (supplemental Figure 6). Of 113 variants, 91 (81%) were concordant between RNA and DNA, 6 (5%) variants were only called on the RNA level, and 16 (14%) variants were only called on the DNA level (supplemental Table 10). Of the 16 variants called on DNA only, 4 were filtered out because of RNA-specific filters (supplemental Methods: RNA-specific filter a-e), and the remaining 12 were not called because of low gene expression and low variant allele frequency.

Large aberrations and small mutations (SNVs and indels) shape the MPN mutational landscape. (A) Distribution of diagnosis within the MPN cohort (outer ring). Occurrence of the MPN driver mutation status within each diagnosis (inner ring). Number of patients with each diagnosis (center). Healthy controls are in gray. (B) Transcriptome sequencing was performed on granulocyte RNA from 113 MPN patients and 15 healthy controls. Independent workflows for fusion calling, variant calling (SNVs and indels), and differential splicing analysis (in SF3B1-mutated patients) were established. Mutations were called, filtered, and validated; those leading to protein changes were translated to amino acid sequences. Neoantigens were predicted on a personalized level, taking each patient’s MHCI haplotype into consideration. (C) A total of 123 MPN patient samples is depicted and sorted by diagnosis, MPN driver mutation status (JAK2, CALR, or MPL positive), and nondriver mutation frequency. Patient sample replicates (n = 10) sequenced across or with the same batch (A-E) are indicated in capital letters (A-J). Chromosomal aberrations panel: uniparental disomies (UPDs), deletions, and gains were called with Affymetrix SNP 6.0 arrays (supplemental Table 5). Fusion panel: fusions were private among patients and colored by rearrangement type. Two fusions were reported for patient P106A#A. For fusion calling, we combined the results of 3 fusion detection tools (deFuse, SOAPfuse, and TopHat-Fusion) to overcome algorithm-specific biases, a practice that is frequently suggested in fusion benchmarking studies.39,40  MPN driver mutation panel: MPN driver mutation status (JAK2, CALR, and MPL) was determined as described in supplemental Methods. SNVs and indels panel: genes are grouped by occurrence in pathways and by mutation frequency. Mutations are colored by mutation type, and a gradient for high and low SIFT score was applied (only for SNVs). Small rectangles with black frame enhance visibility for mutations with low SIFT score. Only those genes with ≥2 mutations across the cohort, with the exception of genes involved in the splicing machinery (SRSF2, SF3A1), are shown. Gene names in blue are part of the TruSight Myeloid Sequencing Panel. For validation, variants from 77 RNA/DNA pairs were compared (supplemental Figure 6). Of 113 variants, 91 (81%) were concordant between RNA and DNA, 6 (5%) variants were only called on the RNA level, and 16 (14%) variants were only called on the DNA level (supplemental Table 10). Of the 16 variants called on DNA only, 4 were filtered out because of RNA-specific filters (supplemental Methods: RNA-specific filter a-e), and the remaining 12 were not called because of low gene expression and low variant allele frequency.

Close Modal
This Feature Is Available To Subscribers Only

or Create an Account

Close Modal
Close Modal