Thread by @VanAllenLab

A thread on our latest look at cancer genomics + immune checkpoint blockade → congrats @diana_miao, Claire, @NIVokes et al!

Thx @SU2C @BroadIgnite @HHMINEWS @theNCI for supporting this & the ppl doing it @NatureGenet @DanaFarber @broadinstitute

rdcu.be/5iVz [1/n]

We analyzed every tumor-normal exome from patients getting immune checkpoint blockade (ICB) we could get & integrated w/ clinical outcomes for biological and clinical exploration → lots of technical pain here + open questions re: defining clinical benefit

For context, lots of ongoing biology re: mutations/neoantigens driving response to immune checkpoint blockade by many groups (including ours).

In parallel, lots of buzz about tumor mutational burden (TMB) as a *clinical* biomarker for cancer ICB...

However, the line btwn biological exploration & clinical biomarker application is getting muddy. Is TMB a singular clinical biomarker in cancers? Here’s ROC curves from @colli_leandro et al 2016 (ncbi.nlm.nih.gov/pubmed/?term=2…) & here’s ours in this cohort.

Err...

(Tangentially, not even touching many technical issues re: measuring TMB across panels, settings, etc. We did QC using new methods here & had to remove many clinical samples. See also: curious result from prospective lung ca TMB study - ncbi.nlm.nih.gov/pubmed/29658845)

So…we took a swing at trying to go beyond TMB in the genomics space and hunt for new somatic genomic features that correlated with ICB benefit and could inform biological discovery (and maybe clinical use)

tl;dr We are woefully underpowered to make any definitive conclusions, and everything we observed needs considerations re: confounders, interacting biology, functional/clinical validation, etc

For example: We & others previously saw PTEN loss (PI3K pathway activation) correlated w/ ICB resistance.

However, here we saw PIK3CA mutation (PI3K pathway activation) correlated w/ ICB *response*…

…but those mutations were in APOBEC signature positive tumors, which itself was correlated w/ IO response…which was also correlated w/ high TMB, which itself correlated w/ response.

Here’s another fun one: TMB correlation with melanoma in aggregate (significant) vs. when you consider mutational signature (not significant) 🤔

Neoantigen biology to study...but perhaps not clearly clinic-ready for this disease at least?

That said, many interesting but underpowered findings here (a few noted in pics), for which we’ll need far more clinical/genomics data AND functional evaluation (i.e. I still have to talk to @nickhaining)…

...and we hope this is another step toward such discoveries.

In retrospect, of course this was going to be hard.

Single gene-drug biomarkers (e.g. BRAF, NTRK) are very effective & still not perfect, so why would therapies that interfere w/ complex tumor-immune system be so simply stratified clinically?

It'll likely take tumor-intrinsic, immune, & clinical features in combination with team science to put this puzzle together. To help, all this data is in @cbioportal for exploration of genomics & response to ICB in this cohort:

cbioportal.org/study?id=mixed…

Onwards, together! [fin]