Number of Patients in COAD Project / Co-occurence

[Silamp]

Hey! I wanna start off by saying that, especially for a beginner like me, maftools is an awesome and easily understandable tool (exactly what I was looking for!) I have a question regarding the patient number of the COAD (colorectal adenocarnioma) TCGA cohort. When I check the firehose website. there are 460 patients in this cohort. (https://gdac.broadinstitute.org). However, when I load the Dataset (I assume it's the firehose dataset) and create the oncoplot, it seems that there are only 406 samples (see attached picture 2). Why is that? Are there mutation annotations missing for some of the initial 460 patients thus resulting in only 406 profiled samples?

coad = tcgaLoad(study = "Coad")
oncoplot(maf = coad, top = 20)

Furthermore, I have another question regarding the co-occurence function. I've used it before to look at a set of genes.

test <- subsetMaf(coad, genes =  c("BRAF", "AXIN2", "CDKN2A", "APC", "CTNNB1", "RNF43", "ZNRF3", "TP53", "SMAD4", "TGFBR1", "TGFBR2", "MLH1", "MSH2", "NOTUM", "PTEN", "KRAS", "PRKCZ", "PRKCZ", "PRKCI", "KRAS", "RSPO2", "RSPO3" ))

somaticInteractions(maf = test, pvalue = c(0.05, 0.1))

The output plot I get is always cut on the sides. Is there a way to avoid this? (See attached picture 1)

Moreover, I assume that this function only tests co-occurence between single genes. Is there a way to look at the co-occurence of multiple genes. For example to answer the question: "If a patient has a mutation in both APC and TP53 does a KRAS mutation co-occur with BOTH APC AND TP53?"

Finally, I was wondering if you've heard of DISCOVER (https://github.com/NKI-CCB/DISCOVER) Its another tool than the FISHER'S EXACT TEST used in Maftools to look at co-occurence. As you know the validity of the FISHERS test depends on the assumption that genes’ alterations across tumors are independent and identically distributed (i.i.d.). Identical distribution implies that the probability of an alteration in a gene is the same for any given tumor. With cancer’s heterogeneity in mind, this assumption may prove problematic. Surely, a gene is more likely found altered in tumors with many somatic alterations overall, than in tumors with only few such changes. The DISCOVER package controls for this issue. What are your thoughts about this and have you thought about implementing this test into maftools?

Sorry for all the questions (I've just started my bioinformatics journey a month ago^^). I am happy for your answers!

Cheers,

silamp

Session info ![oncoplot](https://user-images.githubusercontent.com/101824642/159059263-57e62d33-ca3 ![COOCCURENCE](https://user-images.githubusercontent.com/101824642/159065655-70200cf9-80db-40c3-a5b8-8b801c7217ce

.png) 4-480a-bb06-71aab6637500.png)

Run sessionInfo() and post the output below

[PoisonAlien]

Hi,

Thank you and I am glad that you find the tool user-friendly/helpful.

1. tcgaLoad by default loads a cohort from MC3 project. You will have to specify the source as Firehose.

coad = tcgaLoad(study = "Coad", source = "Firehose")

2. Try decreasing the font size. Maybe fontSize = 0.4?
3. I have not used DISCOVER package but it does look nice. There are more than a few packages to perform co-occurance analysis of pathways (N>2 genes), and maftools just uses fishers test for simplicity for pairwise comparison.

I hope I was able to help. I wish you all the best for the learning curve ahead.

[Silamp]

Hi,

Thank you very much for your answer! Adjusting the fontsize fixed the issue! I've been looking now for quite a while for a co-occurence package for (N>2) genes but I didn't find anything useful. Do you have any recommendations since you mentioned that there are quite a few?

[PoisonAlien]

Check out Raphales group projects . They have multiple approaches. For example CoMEt and WExT.