Genomic, clinical data, and scripts for PD-1 blockade resistance in metastatic melanoma
Abstract
We analyzed whole-exome-sequencing (WES) of pre-treatment tumor and matched normals from four cohorts (n=140) of previously ICB-naïve aPD-1 ICB treated patients. We found high intratumoral genomic heterogeneity and low ploidy robustly identified patients with intrinsic resistance to aPD-1 ICB. Utilizing a melanoma cohort from a period prior to targeted- and ICB-therapy (“untreated” cohort), we found that genomic heterogeneity was predictive while ploidy was prognostic. To establish clinically actionable predictions, we optimized a predictive model using ploidy and heterogeneity to identify, with high confidence (90% PPV), a subset of patients with intrinsic resistance to and worse survival on aPD1 ICB. We validated this model with independent cohorts, and further showed that a significant proportion of patients predicted to have intrinsic resistance to single agent aPD-1 ICB responded to combination ICB, suggesting these patients may benefit disproportionately from combination ICB.
README: aPD1_predictive_model
https://doi.org/10.5061/dryad.nzs7h450g
Description of the data and file structure
Scripts and data necessary to support the research findings.
This repository provides sample code to reproduce figures and all the findings from Tarantino et al. Genomic heterogeneity and ploidy identify patients with intrinsic resistance to PD-1 blockade in metastatic melanoma, with associated source data (supplemental data and tables from the paper).
Please contact the corresponding or the first author for any questions, comments, or concerns regarding the paper in general.
giuseppe_tarantino@dfci.harvard.edu
Files and variables
- Main folder "Validation_def"
subfolder input with the data to reproduce the analysis:
- The supplementary tables of the paper (1-5)
Supplementary table 1 is the discovery cohort, this is the cohort for which we have the most detailed clinical annotation:
- This file contains both:
- Genomic data, columns:
purity - represents the purity of the tumor sample estimated from paired tumor and normal using ABSOLUTE
ploidy - represents the # of complete set of chromosomes of the tumor sample estimated from paired tumor and normal using ABSOLUTE
Gen_w_Aneuploidy - represent the proportion of the genome with copy number alterations
BRAF_status - WT (wild type), MUTANT (mutated)
NRAS_status -WT (wild type), MUTANT (mutated)
n_snvmulti_2 - number of mutations with snv multiplicity of 2
n_snvmulti_1 - number of mutations with snv multiplicity of 1
ratio_snvmulti - ratio n_snvmulti_2 to n_snvmulti_1
loh - loss of heterozygosity estimated from paired tumor and normal using ABSOLUTE
ratio_loh - proportion of loh
snv_m_score - the formula is described in the paper "SNV multiplicity score=WGD+WGD*SNV multiplicity 2to1 ratio"
heterogeneity - proportion of subclonal mutations over the total non synonymous mutations
#clonal - number of clonal mutations
#subclon - number of clonal mutations
total_nonsyn - total number of non synonymous mutations
* Clinical data, columns:
* BR (Best response) to the treatment, can be Stable disease (SD), progressive disease (PD), Complete response (CR), Partial response (PR), Mixed response (MR)PD (Progressive disease) and PD_cat - 1=yes and 0=not
RvsP 3=SD or MR, 2=PD, 1=CR
(Right Ventricular Systolic Pressure)
Cohort, can be Schadendorf (from Liu et al. Nature Medicine 2019), bms_checkmate_064 (from the clinical trial checkmate 064), and bms_checkmate_038 (from the clinical trial checkmate038)
os_days - Overall survival days
dead - 1=death , 0=censored
pfs_days - Progression free survival days
progressed - 1=progressed , 0=censored
GD - Genome doubling, 1=Yes, 0=no
ECOG - Eastern Cooperative Oncology Group (ECOG) performance status, 0=Fully active, able to carry on all pre-disease performance without restriction; 1=Restricted in physically strenuous activity but ambulatory and able to carry out work of a light or sedentary nature.
Primary_type - Melanoma primary type: Skin, Mucosal, Other, Ocular/Uveal, Acral, occult
Mstage - Melanoma Metastasis staging: M0, M1A, M1B, M1C
Brain_met - presence of brain metastasis 0=no, 1=yes
LDH_cat - LDH levels 0=normal, 1=abnormal
Age - Age at diagnosis
Sex - Male or Female
Liver_Met - presence of liver metastasis 0=no, 1=yes
Lung_Met - presence of lung metastasis 0=no, 1=yes
prior_BRAFi - prior BRAF therapy
TMB - tumor mutational burden estimated from whole exome sequencing
ploidy_q - ploidy quartile groups
het_q - heterogeneity quartile groups
TMB_q - tumor mutational burden quartile groups
* Models predictionslogreg_pred - prediction from logistic regression model, 1=predicted as PD, 0= predicted as NPD
DT_pred - prediction from decision tree model, 1=predicted as PD, 0= predicted as NPD
DT_mod_pred - prediction from modified decision tree model to optimize precision, 1=predicted as PD, 0= predicted as NPD
Supplementary table 3 the list of features tested
- column Signature/model refers to the feature or the model that was tested
- column Original Study is the reference for that specific signature or model
Supplementary table 5 the combo ICB cohort
- Genomic data, columns:
purity - represents the purity of the tumor sample estimated from paired tumor and normal using ABSOLUTE
ploidy - represents the # of complete set of chromosomes of the tumor sample estimated from paired tumor and normal using ABSOLUTE
heterogeneity - proportion of subclonal mutations over the total non synonymous mutations
#clonal - number of clonal mutations
#subclon - number of clonal mutations
total_nonsyn - total number of non synonymous mutations
* Clinical data, columns:OS_days - Overall survival days
OS_event - 1=death , 0=censored
PFS_days - Progression free survival days
PFS_event - 1=progressed , 0=censored
DCB (durable clinical benefit)
PD (Progressive disease) based on PFS; 1 means PFS<=6 months ; 0 PFS>6 months
prior_cpi_or_v - prior immune checkpoint blockade
missing data: NA
TCGA melanoma, the data from TCGA, specifically reanalyzed data have been obtained JR. Conway et al. from https://doi.org/10.1038/s41588-020-00739-1
TotalCN_circos to reproduce the circos plots
- subfolder for the progressor and responders
- the input files are the "*_processed.tsvcircos"
- the columns are chromosome, segment start, segment end, color
- the configuration file is the conf_classici.txt
- file with the dimension for each circos
subfolder *scripts *with the code to reproduce the main figures
- Figure1.Rmd for the first figure
- Figure2.Rmd for the second figure
- Figure3.Rmd for the third figure
- Figure4.Rmd for the fourth figure
- Figure5.Rmd for the fifth figure
- useful_functions.R functions used to plot confusion matrix and decision boundaries
Code/software
the version used are specified in each script
Access information
Other publicly accessible locations of the data:
Data was derived from the following sources:
- R. Conway et al. from https://doi.org/10.1038/s41588-020-00739-1
- S. Freeman et al from https://doi.org/10.1016/j.xcrm.2021.100500
Methods
Metastatic melanoma patients treated with immune checkpoint blockade were identified from published work (Liu et al. Nature Medicine 2019 & Freeman et al. Cell Reports Medicine 2022) and completed clinical trials (BMS Checkmate 038 and checkmate 064). We included only samples without prior exposure to ipilimumab, with WES data of the paired tumor and normal tissue obtained before PD1 blockade. Clinicopathological and demographic data were obtained from Liu et al Nature Medicine 2019, from BMS for the two clinical trials and for the validation cohort from Freeman et al. Data are shown in Fig. 1 and in Supplementary table 1. The best objective response (BOR) to aPD1 ICB was only available for a subgroup of the patients included in Freeman et al. and wasn’t available for the combination immunotherapy-treated (“combo”) cohort.
Samples from the BMS and Freeman et al. cohorts were re-analyzed with the Broad Institute CGA pipeline (57–67) using the TERRA platform, adopting the same quality controls filters used for the Liu et al. Nature Medicine 2019. In particular quality control cutoffs were as follows: mean target coverage > 50X (tumor) and >30X (normal), cross contamination of samples estimation (ContEst)<5%, tumor purity >= 10%, DeTiN ≤ 20% TiN. A power filter combining coverage and tumor purity was applied as described (e.g. minimum 80% power to detect clonal mutations) in Liu et al. Nature Medicine 2019. Three samples were excluded for low purity and two samples for low power.
MuTect2 was used to identify somatic single-nucleotide variants in targeted exons, with computational filtering of artifacts introduced by DNA oxidation during sequencing or FFPE-based DNA extraction using a filter-based method. Subsequently Strelka was used to identify small insertions or deletions. Lastly, Oncotator was used to annotate the Identified alterations.
Absolute was used for the estimation of ploidy, and purity and for the cancer cell fraction (CCF) estimation of individual mutations. For each sample, the optimal solution (purity, ploidy) was manually selected among the local solutions. Heterogeneity was computed as the proportion of the subclonal mutations, with a mutation defined as subclonal if the cancer cell fraction (CCF) was lower than 0.8. To support the cutoff of 0.8 for CCF, we have performed a sensitivity analysis (supplementary fig. 4), demonstrating that the heterogeneity stratification was maintained even when different cutoff values were utilized.