TMB and response to ICB

Association of TMB with response to ICB across five cancer types from CPI800+ (the largest cohorts of each cancer type are plotted here, the others are shown in Figure S1A). Only melanoma and non-small cell lung cancer have a significantly different TMB between responders and non-responders. ccRCC: clear cell Renal Cell Carcinoma, HNSCC: Head and Neck Squamous Cell Carcinoma; NSCLC: Non Small Cell Lung Carcinoma

TMB association with progression-free survival post-immunotherapy

Plots of progression-free survival and TMB for melanoma and lung cancer ICB cohorts show the lack of correlation or of an obvious TMB cutoff.

TMB as a biomarker of response to immunotherapy (A)(B)

ROC curves for the melanoma and lung cancer cohorts. Youden index associated cutoffs are also plotted. (C) Boxplots of nonsynonymous mutation rates across responders and non responders in the melanoma and lung cancer cohorts. The FDA-approved cutoff (10 mutations/Mb) and the best cutoff (Youden index associated cutoff) are shown by vertical lines. (D) Proportion of misclassified patients based on the FDA-approved cutoff, as well as the Youden index cutoff for each dataset. The use of either cutoff leads to substantial fraction of misclassified patients (potential responders below the treatment cutoff, or non-responders above the cutoff).

TMB association with progression-free survival post-immunotherapy

(A) Overview of the randomization analysis. Left: the optimal cutoff is found to maximize the difference between survival between groups above and below the cutoff (i.e to minimize the logrank p-value, yielding preal). Right: the same procedure for shuffled data yields pshuf. The fraction of pshuf<preal produces a p-value corrected for multiple hypothesis testing for non-independent tests. (B) Results of the randomization analysis in the melanoma cohorts and stratification by subtypes (p-values < 10-10 not shown) (C) Randomization analysis results in the lung cancer cohorts and stratification by subtypes (p-values < 10-10 not shown).

TMB and cancer immunogenicity (A)

Our model of cancer immunogenicity coarse-grains several cellular processes into the probability that a mutation becomes immunogenic (Pimmunogenic). If the number of immunogenic mutations reaches kcrit, the cancer triggers an immune response (B) The probability of immune response Pimmune responce as a function of TMB for a range of kcrit and Pimmunogenic. Rapid saturation of Pimmune responce. TMB requires low kcrit and sufficiently high Pimmunogenic>0.1 (see Materials and Methods).

Figure 5B and Figure S7 show P{immune response} as a function of N (TMB) for a range of p and kcrit values. The model shows two different behaviors. If individual mutations are unlikely to be immunogenic p ≪ 1, e.g. due to a low probability of being presented,, the probability of response increases gradually with TMB. The neoantigen theory generally expects such gradual increase in immunogenicity of cancer with TMB. Yet, available data (Figure 2) don’t show such a trend.

TMB association with clinical benefit from ICB across cancers (A)

Association of TMB with response to ICB across three cancer types from CPI800+. Only melanoma and non-small cell lung cancer have a significantly different TMB between responders and non-responders. (B) Association of TMB with response to ICB across seven cancer types from CPI1000+. Melanoma, non-small cell lung, bladder and colorectal cancer have a significantly different TMB between responders and non-responders.

TMB as a biomarker of response to immunotherapy

ROC curve for CPI1000+. The Youden index associated cutoffs is also plotted.

TMB association with progression-free survival post-immunotherapy

(A) (B) Plots of progression-free survival and TMB for melanoma and lung cancer ICB cohorts labeled by cancer subtype, showing the lack of correlation or of an obvious TMB cutoff.

TMB association with overall survival post-immunotherapy

Results of the randomization analysis in CPI1000+ (p-values < 10-10 not shown).When cancer types of CPI1000+ were combined, a nominally significant p-value (p=0.04) arises, likely due to cancer types with different TMB ranges showing significantly different survival rates to ICB.

TMB association with overall survival post-immunotherapy (A)

Randomization analysis results in mel1 and mel2 and stratification by subtypes (p-values < 10-10 not shown) (B) Randomization analysis results in lung1 and stratification by subtypes (p-values < 10-10 not shown). When corrected for multiple hypotheses all cohorts fail to provide a statistically significant cutoff.

TMB association with overall survival post-immunotherapy

Randomization analysis results in multiple cancer types with MSK-IMPACT targeted next-generation sequencing data (p-values < 10-10 not shown)

Components of cancer immunogenicity

(A) Probability of eliciting an immune response for a range of kcrit values (B) Probability of eliciting an immune response for a range of Pimmunogenic values