I am noticing low tumour purity in some of my samples based on the orange report. Briefly discussed this here before, noting that low purity does not always equate to low quality overall, as variants are still found. But is there anything I can change when running oncoanalyser to improve the purity / QC scores? These samples are tumour only, from FFPE tissue blocks and some are a few years old so Im guessing there is not much that can be done

In a clinical setting we fail samples below 30% purity. You can counteract that to some extent by sequencing really deeply but you will miss a lot of relevant information either way.

Thanks @Oliver Hofmann They attempted to sequence deeply, at ~90x coverage. However most samples are coming back at mean depth ~30-60.

For FFPE we aim for ~120 raw coverage which ends up being ~70x after deduplication, but yeah, it can be tricky. Old FFPE blocks in general will give you lots of noise.

Why does it say Purity: 8% (50-100%)? What does that range in brackets mean?

It's the spread of purity values among other candidate solutions in PURPLE's grid search that scored almost as well as the single best-scoring one - there's some "confidence" in that the smaller the range, the more confident PURPLE is

So the spread was between 50 and 100%, but it thought 8% was the most likely?

The meaningfulness of that number will highly depend on the type of somatic fitting that was done (based on copy numbers or somatic, variants - depending on the numbers present) but it's a known challenge with FFPE samples, tumor-only on top of that

Ok, it just seemed like a bug to be honest

Quite challenging to do this with such samples 😅

well, you'd expect it to say (8-100%)

Feel free to make an issue here

Hi @Rayan Hassaïne Can you elaborate on "It's worked out in the upcoming version of the pipeline" Will there be some improvements to the purity estimation? When is this version expected to be released?

yes there are whole set of improvements on all fronts and we're putting the finishing touches on for the release!

Hi @Rayan Hassaïne Approximately when will the new version of oncoanalyser be released? I have new samples to analyse and am considering waiting for this newer version. Thanks

I am going to say any day now (and then additional time for the nf-core review and addressing potential concerns) but don't want to add more stress to the team 🙂