Tumor mutational landscape is a record of the pre-malignant state

Kübler, Kirsten; Haradhvala, Nicholas J.; Ha, Kyungsik; Kim, Jaegil; Kuzman, Maja; Jiao, Wei; Gakkhar, Sitanshu; Mouw, Kent W.; Braunstein, Lior Z.; Elemento, Olivier; Biankin, Andrew V.; Rooman, Ilse; Miller, Mendy; Karthaus, Wouter R.; Nogiec, Christopher D.; Juvenson, Edouard; Curry, Edward; Mino‐Kenudson, Mari; Ellisen, Leif W.; Brown, Robert V.; Gusev, Alexander; Tomasetti, Cristian; Lolkema, Martijn P.; Steeghs, Neeltje; Herpen, Carla van; Kim, Hong-Gee; Lee, Hwajin; Vlahoviček, Kristian; Bernstein, B; Sawyers, Charles L.; Hoadley, Katherine A.; Cuppen, Edwin; Koren, Amnon; Arndt, Peter F.; Louis, David N.; Stein, Lincoln; Foulkes, William D.; Polak, Paz; Getz, Gad

doi:10.1101/517565

Cited by 23 publications

(30 citation statements)

References 92 publications

(102 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For example, ovarian cancers are distinguished by a high rate of genomic rearrangements 8 , chronic myelogenous leukaemias (CML) carry a nearly pathognomonic structural variation involving a t(9;22) translocation leading to a BCR-ABL fusion transcript 9 , melanomas have high rates of C > T and G > A transition mutations due to UV damage 10 and pancreatic ductal adenocarcinomas have near-universal activating mutations in the KRAS gene 11 . Recent work has pointed to a strong correlation between the regional somatic mutation rate and chromatin accessibility as measured by DNase I sensitivity and histone mark 12 , and has suggested that the cell of origin can be inferred from regional mutation counts 13 .…”

mentioning

confidence: 99%

A deep learning system accurately classifies primary and metastatic cancers using passenger mutation patterns

et al. 2020

Self Cite

View full text Add to dashboard Cite

In cancer, the primary tumour's organ of origin and histopathology are the strongest determinants of its clinical behaviour, but in 3% of cases a patient presents with a metastatic tumour and no obvious primary. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, we train a deep learning classifier to predict cancer type based on patterns of somatic passenger mutations detected in whole genome sequencing (WGS) of 2606 tumours representing 24 common cancer types produced by the PCAWG Consortium. Our classifier achieves an accuracy of 91% on held-out tumor samples and 88% and 83% respectively on independent primary and metastatic samples, roughly double the accuracy of trained pathologists when presented with a metastatic tumour without knowledge of the primary. Surprisingly, adding information on driver mutations reduced accuracy. Our results have clinical applicability, underscore how patterns of somatic passenger mutations encode the state of the cell of origin, and can inform future strategies to detect the source of circulating tumour DNA.

show abstract

mentioning

confidence: 99%

A deep learning system accurately classifies primary and metastatic cancers using passenger mutation patterns

et al. 2020

Self Cite

View full text Add to dashboard Cite

show abstract

“…[1][2][3][4] For example, ovarian cancers are distinguished by a high rate of genomic rearrangements, 5 chronic myelogenous leukemias (CML) carry a nearly pathognomonic structural variation involving a t(9;22) translocation leading to a BCR-ABL fusion transcript, 6 melanomas have high rates of C>T and G>A transition mutations due to UV damage, 7 and pancreatic ductal adenocarcinomas have near-universal activating mutations in the KRAS gene. 8 Recent work has pointed to a strong correlation between the regional somatic mutation rate and chromatin accessibility as measured by DNase I sensitivity and histone mark 9 , and has suggested that the cell of origin can be inferred from regional mutation counts 10 .…”

Section: Introductionmentioning

confidence: 99%

A deep learning system can accurately classify primary and metastatic cancers based on patterns of passenger mutations

Jiao

Atwal

Polak

et al. 2017

Preprint

Self Cite

View full text Add to dashboard Cite

In cancer, the primary tumour's organ of origin and histopathology are the strongest determinants of its clinical behaviour, but in 3% of the time a cancer patient presents with metastatic tumour and no obvious primary. Challenges also arise when distinguishing a metastatic recurrence of a previously treated cancer from the emergence of a new one. Here we train a deep learning classifier to predict cancer type based on patterns of somatic passenger mutations detected in whole genome sequencing (WGS) of 2606 tumours representing 24 common cancer types. Our classifier achieves an accuracy of 91% on held-out tumor samples and 82% and 85% respectively on independent primary and metastatic samples, roughly double the accuracy of trained pathologists when presented with a metastatic tumour without knowledge of the primary. Surprisingly, adding information on driver mutations reduced classifier accuracy. Our results have immediate clinical applicability, underscoring how patterns of somatic passenger mutations encode the state of the cell of origin, and can inform future strategies to detect the source of cell-free circulating tumour DNA.

show abstract

“…Since the immune signature in the tumor microenvironment is highly correlated with the response to treatments such as immunotherapy, germline variants could enhance predictive modeling of response and reveal novel therapeutic targets for shifting infiltration profiles to a more favorable one. Previous studies demonstrated that cancer cells maintain chromatin structure from the tissue-of-origin, so it is possible that germline iQTLs have conserved infiltration effects in the cancer cells 53 . If this were the case, then functional experiments could be a promising avenue for developing medicines to shift infiltration patterns.…”

Section: Discussionmentioning

confidence: 99%

Age, Sex, and Genetics Influence the Abundance of Infiltrating Immune Cells in Human Tissues

Marderstein

Uppal

Verma

et al. 2019

Preprint

Self Cite

View full text Add to dashboard Cite

Despite infiltrating immune cells playing an essential role in human disease and the patient response to treatment, the central mechanisms influencing variability in infiltration patterns are unclear. Using bulk RNA-seq data from 53 GTEx tissues, we applied cell-type deconvolution algorithms to evaluate the immune landscape across the healthy human body. We first performed a differential expression analysis of inflamed versus non-inflamed samples to identify essential pathways and regulators of infiltration. Next, we found 21 of 73 infiltration-related phenotypes to be associated with either age or sex (FDR < 0.1). Through our genetic analysis, we discovered 13 infiltration-related phenotypes have genome-wide significant associations (iQTLs) (P < 5.0 x 10 -8 ), with a significant enrichment of tissue-specific expression quantitative trait loci in suggested iQTLs (P < 10 -5 ). We highlight an association between neutrophil content in lung tissue and a variant near the CUX1 transcription factor gene (P = 9.7 x 10 -11 ), which has been previously linked to neutrophil infiltration, inflammatory mechanisms, and the regulation of several immune response genes. Together, our results identify key factors influencing interindividual variability of specific tissue infiltration patterns, which could provide insights on therapeutic targets for shifting infiltration profiles to a more favorable one. Results Robust estimation of immune cell types in bulk RNA-seq profiles.To describe immune content from bulk RNA-seq samples, we used two central algorithms: xCell 13 and CIBERSORT 14 . xCell relies on a modification of single sample gene-set enrichment analysis to estimate cell type scores, while CIBERSORT employs a linear support vector regression model. The default reference signatures allow deconvolution of 64 immune and stroma cell types for xCell and 22 immune cell types for CIBERSORT. CIBERSORT also calculates a scaling factor that measures the degree of infiltration. We refer to the relative proportions from CIBERSORT as "CIBERSORT-Relative" and the product of the relative proportions with the scaling factor as "CIBERSORT-Absolute". We estimate three scores for each cell type to describe the immune content from the gene expression data for each tissue in each individual: xCell, CIBERSORT-Relative, and CIBERSORT-Absolute scores.We first hypothesized that the relative and absolute scores from CIBERSORT encapsulated different aspects of the single-cell deconvolution. While "CIBERSORT-Absolute" simultaneously quantifies a degree of immune infiltration, "CIBERSORT-Relative" is focused on capturing compositional changes in the immune content (Supplementary Note). We simulated synthetic mixes composed of bulk tissue "spiked" in silico with CD4+ T cells and CD8+ T cells (see Methods). We correlated the known amount of CD4+ and CD8+ T cell infiltration in these mixtures with estimated deconvolution scores under a "tissue" scenario and an "immune cell" scenario. In the "immune cell" scenario, we let the true infiltration be the prop...

show abstract

Tumor mutational landscape is a record of the pre-malignant state

Cited by 23 publications

References 92 publications

A deep learning system accurately classifies primary and metastatic cancers using passenger mutation patterns

A deep learning system accurately classifies primary and metastatic cancers using passenger mutation patterns

A deep learning system can accurately classify primary and metastatic cancers based on patterns of passenger mutations

Age, Sex, and Genetics Influence the Abundance of Infiltrating Immune Cells in Human Tissues

Contact Info

Product

Resources

About