Fancy a challenge? A DREAM of intra-tumour phylogenies

You are into tumor evolution? And got a fancy model? Want to battle with the best?

Then check out the ICGC-TCGA DREAM Somatic Mutation Calling – Tumour Heterogeneity Challenge (SMC-Het).

These are the days of Big Science, my friend. You can’t just have a short name …

The ICGC-TCGA DREAM Somatic Mutation Calling – Tumour Heterogeneity Challenge (SMC-Het) is an international effort to improve standard methods for subclonal reconstruction: to quantify and genotype each individual cell population present within a tumor. Leaders of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA) cancer genomics projects are joining with Sage Bionetworks and IBM-DREAM to initiate this innovative open crowd-sourced Challenge [1-3].

The goal of this challenge is to identify the best subclonal reconstruction algorithms and to identify the conditions that affect their performance. By running an unbiased, comprehensive evaluation, the most appropriate method(s) will be identified.

There is not much info yet on how the data will look like. But the challenges are exactly what you would expect them to be:

Subchallenge 1: Build a model that best estimates the number of subclones with their proportion.

Subchallenge 2: Identify the most predictive methods to predict lineage of evolutionary tree. Given the subclonal reconstruction(s) algorithms, the aim is to compare them to a known true structure (“is-ancestor-of”, “is-descendant-of”, “co-clustering” and “is-sibling/cousin”).

I pre-registered as soon as I saw the challenge, which was today. Pre-registration opened November 10, but it can’t have been widely advertised, because I completely missed it.

Now I am excited like a 5-year old the day before Christmas.

What will they give us?

Deep-sequencing data?

SNVs?

Structural variation?

Single cell data?

Simulated data?

A mix of the above?

And how can you estimate sub-clones (challenge 1) without having the tree (challenge 2)? Remember, that clusters of mutations are not yet clones.

How will the results be validated?

How did they derive the gold standard solution?

What metrics will be used to compare different methods?

Soooo many questions.

Florian