Science - USA (2022-04-22)

INSIGHTS | PERSPECTIVES

GRAPHIC: K. FRANKLIN/

SCIENCE

science.org SCIENCE

study of Degasperi et al. also provides fertile
ground for further investigations.
The mechanisms that lead to muta-
tional signatures are most often validated
or elucidated in cultured cells, using envi-
ronmental mutagenic treatments or geneti-
cally modified cell lines (8 –10). Such experi-
ments can provide clear mutation spectra,
but causative connections must be drawn
carefully because different processes may
produce very similar spectra. For a bet-
ter understanding, investigators may also
focus on attributes beyond the immediate
sequence context, such as transcription or
replication strand bias (which were also re-
ported by Degasperi et al.), clustering, as-
sociations with replication timing, wider
sequence context, and chromatin features.
Such information can be handled separately
or incorporated into signatures ( 11 ), with
the latter approach complicated by multi-
dimensionality, or a need to assign relative
weights to the components. The number of

mutations in individual samples also limits

the complexity of analyses.

The primary incentive of cancer muta-

tional signature analysis is to provide diag-

nostic tools for patient stratification ( 12 ).

The first requirement is the fitting of sig-

natures to mutational spectra of individual

samples. Degasperi et al. provide a method

that is analogous to signature decomposi-

tion: Common signatures are fitted first,

followed by the addition of rare signa-

tures one at a time. Fitting organ-specific

signatures defined on the largest datasets

is likely to be particularly accurate, and

it may also help determine the organ of

origin of metastatic samples. The scale of

the analysis demonstrated that multiple

organs have rare cases of tumors with sig-

natures that are common in other organs,

for example, those attributed to deficiency

of DNA repair through homologous recom-

bination or mismatch repair, which may

benefit from targeted treatment.

CNS, Central nervous system

Samples with rare signatures

Samples with common signatures only

Tumor samples

Two-step tting

of signatures

to individual

tumor samples

Organ-specific rare mutational signaturesNew

New

Organ-specific common mutational signatures

Mutational spectra of tumors from a single

organ are split into component signatures

Bone and soft tissue Breast CNS Colorectal Kidney Lung Ovary Uterus

102

101

103

104

105

106

Base substitutionsper tumor genome

Reliable information is needed on the

predictive value of mutational signatures

for drug sensitivities, which is most likely

to find use in the treatment of DNA re-

pair–deficient tumors. Many samples

with DNA repair deficiency–associated

signatures lack identifiable driver gene

mutations because of an incomplete un-

derstanding of the underlying biology or

chromatin-dependent gene expression

changes, arguing for the preferred use of

mutational signature analysis rather than

targeted gene sequencing. Indeed, a pre-

liminary study reports that the association

of drug sensitivities with mutational sig-

natures is stronger than that with driver

mutations ( 13 ). However, both in vitro and

clinical investigations have found that the

expected correlation between mutational

signatures connected with homologous

recombination deficiency and sensitivity

to poly(ADP-ribose) polymerase (PARP)

inhibitors (which target this defect) is less

than straightforward ( 14 , 15 ).

The use of compound classi-

fiers that rely on multiple mu-

tation types may help provide

more accurate predictions.

It must also be remembered

that mutational spectra give

a compressed view of the tu-

mor’s past, and the observed

mutational processes may no

longer operate at the time of

treatment. j

REFERENCES AND NOTES

1. A. Degasperi et al., Science 376 ,
   eabl9283 (2022).


2. S. Nik-Zainal et al., Cell 149 , 979 (2012).


3. L. B. Alexandrov et al., Nature 500 , 415
   (2013).


4. L. B. Alexandrov et al., Nature 578 , 94
   (2020).


5. COSMIC, Mutational signatures (v3.2
   
   
   • March 2021); http://cancer.sanger.
     ac.uk/cosmic/signatures.
   
   
   
   


6. C. Turnbull, Ann. Oncol. 29 , 784 (2018).


7. M. A. Sanders et al., Blood 132 , 1526
   (2018).


8. J. E. Kucab et al., Cell 177 , 821 (2019).


9. X. Zo u et al., Nat. Commun. 9 , 174 4
   (2018).


10. J. Zámborszky et al., Oncogene 36 , 74 6
    (2017).


11. H. Vöhringer et al., Nat. Commun. 12 ,
    3628 (2021).


12. S. W. Brady et al., Trends Genet. 38 , 194
    (2022).


13. J. Levatić et al., bioRxiv
    2021.2005.2019.444811 (2021).


14. Á. Póti et al., Genome Biol. 20 , 240
    (2019).


15. N. Chopra et al., Nat. Commun. 11 , 2662
    (2020).

ACKNOWLEDGMENTS

I thank E. Németh for help with the figure.

10.1126/science.abo7425

352 22 APRIL 2022 • VOL 376 ISSUE 6591

Mutational signature analysis in cancer
Whole-genome sequences of 12,222 tumor-normal sample pairs from the 100,000 Genomes Project of Genomics England
reveal single-base substitutions in different cancer types (top). These mutation spectra were analyzed by Degasperi et al.
to define component mutational signatures (bottom), which are induced by specific mechanisms (many of which are unknown).
These mutational signatures give information about the history of mutagenesis and properties of individual tumors.