Thread by @SerenaNikZainal

Our paper is out! 😊
Showcasing the complexity of #mutationalsignatures

A practical framework and online tool for mutational signature analyses show intertissue variation and driver dependencies ⁦@NatureCancer⁩
👇
1/ nature.com/articles/s4301…

We compare and contrast components of analytical steps in signatures analysis and suggest a practical framework for seeking Mutational signatures in tens to hundreds of tumour samples.

2/

Applying these methods on 3107 WGS primary cancers of 21 organs reveals known signatures and nine previously undescribed rearrangement signatures.
Spot the workflow 👇
3/

We cluster all signatures identified in each tissue type, call a cluster average a RefSig, & highlight variability in distribution of cosine similarities between signatures extracted from different organs and the RefSig.
They don’t behave in a uniform way!
4/

We present a way of visualising that inter-organ diversity.
Try the browsable map here:
signal.mutationalsignatures.com/explore/cancer…

5/

We reinforce our findings of inter-organ variability in an independent analysis of 3096 WGS metastatic cancers.

6/

We look 4 relationships between drivers (A) & signatures (B) in these “target plots”

7/

Only two signatures show a one-to-one relationship between signature and driver 👇👇

Note the semicircles are both dark blue.
8/

Expected relationships found.

Signatures associated with BRCA1/BRCA2 altogether are specific.
Semicircles in left are dark blue.
Individual signatures, e.g. signature 3, may not be very specific....
9/

Some associations are simply associations. Like for these two, the alleged driver has the pattern of the signature and is most likely a consequence of the signature
10/

Tandem duplications come in a variety of sizes and have different biological associations.
11/

Signatures with high genomic instability that can form hypermutator phenotypes are dependent on TP53 dysregulation.

Unlike all other analyses, the blue semicircle is dark on the right in these target plots! (not seen for all signatures with TP53..)
12/

Using Sig 3 alone for BRCAness classification is not great. It gives a high false pos rate. Better to use multiple signature approach like HRDetect.
Essential for accurate tumour stratification.

13/

Check out our web-based tool where you can explore these analyses, including all experimentally-generated data from our group.
And perform signature assignments on your samples!
@SignalCambridge

signal.mutationalsignatures.com

14/

DATA: We r grateful 🙏to #BASIS, ICGC #PCAWG project & #Hartwig for data access.

Code: github.com/Nik-Zainal-Gro…

Special Kudos 👍💪 to #AndreaDegasperi #ScottShooter #JanCzarnecki for driving this project.

Though the whole team👇
were involved really!

/END