Pith. sign in

REVIEW 4 major objections 5 minor 34 references

Bayesian Non-Negative Matrix Factorization with Correlated Mutation Type Probabilities for Mutational Signatures

T0 review · 4 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Covariance-aware NMF recovers more mutational signatures

desk verdict A mostly sound Bayesian NMF extension whose advertised accuracy gain rests on thin, under-specified simulations. read the letter →

arxiv 2506.15855 v2 pith:JVCL2232 submitted 2025-06-18 q-bio.QM stat.ME

Iris Lang , Jenna Landy , Giovanni Parmigiani This is my paper
classification q-bio.QMstat.ME MSC 62F1562P10
keywords mutationalsignaturesBayesianNMFmultivariatetruncatednormalcovariancestructureGibbssamplinghierarchicalmodelsomaticmutationscancergenomics
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Mutational signature analysis typically decomposes a tumor's 96-type mutation counts into a product of signatures and exposures, and standard Bayesian treatments assume the 96 mutation probabilities are independent. This paper argues that mutation types are measurably dependent—mutations sharing a center base are positively correlated—and that ignoring this dependence costs accuracy when samples are few. The authors propose two Bayesian NMF extensions: one that places a Multivariate Truncated Normal prior on the signatures matrix using an externally estimated covariance, and a hierarchical version that learns the covariance itself as two correlation parameters. In simulations, the covariance-aware model reaches the same posterior quality in fewer MCMC iterations and, on small samples, recovers six of eight true signatures where the independent model recovers four. If the claims hold, small cancer cohorts could yield more accurate signatures, and the learned covariance could expose biological connections among mutation types.

What carries the argument

The load-bearing object is the 96-by-96 covariance matrix $\Sigma$ of the signatures matrix $P$. It is inserted as the scale matrix of a Multivariate Truncated Normal prior on each column of $P$, and the Normal likelihood is conjugate enough that the full conditional of each column is again a truncated multivariate normal with precision $\Sigma^{-1} + (\sum_g E_{ng}^2)(\Sigma^M)^{-1}$. In the hierarchical model $\Sigma$ is parameterized as $\sigma_P^2$ times a two-correlation exchangeable structure: $\rho_{\text{same}}$ for mutation types sharing a center base, $\rho_{\text{diff}}$ otherwise, giving a parsimonious, interpretable covariance that can be learned by Metropolis-Hastings within the Gibbs sampler.

What would settle it

Generate a test set from independent truncated normals (identity covariance) and run both models: if the covariance-aware model still recovers more signatures than the independent model, its advantage does not come from modelling true dependence; if it recovers fewer, the in-sample advantage of a matched prior is confirmed.

Watch

Extended reading notes

Core claim

The central claim is that replacing independent truncated-normal priors on each entry of the signatures matrix with a single Multivariate Truncated Normal prior—whose covariance encodes how mutation types co-vary—improves both the speed and the accuracy of Bayesian NMF for mutational signatures. The paper derives the full conditional for each signature column, which remains a truncated multivariate normal, and validates the model in simulations: with sixty samples and five true signatures it converges in hundreds to thousands fewer iterations than the independent model at equal accuracy, and with ten samples and eight true signatures it discovers more of them (six versus four by the paper's cosine-similarity threshold). The hierarchical version replaces the fixed covariance with a three-parameter structure—a base variance and two correlation parameters, one for mutation types sharing the same center base and one for those that do not—and shows on simulated data that the algorithm recovers the generating correlation parameters and, on a pancreas-endocrine cancer cohort, estimates near-zero correlations while still matching reference signatures as well as or better than the independent model.

Load-bearing premise

The simulations that show the accuracy gain generate data from a covariance structure and hand the same structure to the model as its prior, so the reported improvement is an in-sample test; in real applications the external covariance may not match the tumor type being analyzed.

Editorial extensions

If this is right

  • Small sample cohorts (few samples per signature) can recover more true signatures when the prior accounts for mutation-type correlations.
  • The correlated model reaches the same MAP posterior in fewer MCMC iterations, though each iteration costs more, so wall-clock gains depend on implementation.
  • The hierarchical covariance parameters give a compact summary of how mutation types co-vary across cancers, with $\rho_{\text{same}}$ versus $\rho_{\text{diff}}$ interpretable as biological coupling.
  • Because only the prior changes, the formulation carries over to other NMF alphabets and to any non-negative matrix decomposition where columns are expected to be correlated.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the in-sample matched covariance is the source of the gain, then the method's practical value depends on how transferable a reference-catalog covariance is across cancer types; this is directly testable by applying the MVN prior built from one cancer type's signatures to another.
  • The near-zero $\rho_{\text{same}}$ and $\rho_{\text{diff}}$ on the real dataset could mean the simple two-parameter covariance is too restrictive, or that the biological dependence is real but weaker than the reference catalog suggests; distinguishing these would require checking whether a richer covariance (e.g., per-center $\rho$) improves real-data recovery.
  • The paper's convergence criterion is based on stabilization of the MAP, so 'fewer iterations to convergence' measures optimization progress rather than full posterior mixing; a user who needs credible intervals should also check trace convergence of the covariance parameters themselves.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes two Bayesian NMF extensions for mutational signature analysis: (i) a multivariate truncated normal prior on the columns of the signatures matrix, with covariance estimated from COSMIC, and (ii) a hierarchical model in which a three-parameter covariance structure (sigma2_P, rho_same, rho_diff) is learned via Metropolis-Hastings. The authors derive the relevant full conditionals, implement the methods in an open-source R package, and report simulation studies plus an application to Panc-Endocrine PCAWG data. The abstract claims that the MVN model converges in fewer MCMC iterations and improves accuracy, especially on small sample sizes.

Significance. The full conditional derivation in Appendix 7.4 is algebraically sound, and the biological motivation for modeling dependence across mutation types is reasonable. The contribution of open-source code is also useful. However, the main empirical evidence for the headline accuracy claim is an in-sample comparison in which the MVN method is supplied with the generative covariance, and the supporting statistical tests contain methodological flaws. As presented, the paper does not demonstrate that an external or learned covariance improves signature recovery in realistic settings. The significance is therefore currently limited to a proof-of-concept, pending a properly designed evaluation.

major comments (4)
  1. [§4.3 (Figure 6)] The small-sample accuracy comparison is an oracle comparison. The two datasets are generated from the three-parameter covariance with sigma2_P=7, rho_same=0.5, rho_diff=-0.1, and the MVN method is given the same structure as its prior, while the nonMVN method is restricted to an identity covariance. The reported 6/8 versus 4/8 recovered signatures therefore measures the gain from using the true generative covariance against a deliberately misspecified model, not the practical value of a COSMIC-derived or learned covariance. The sentence in Section 4.3, 'Given that this dataset was constructed under the assumption of the 3-parameter hyperprior structure ...', makes this explicit. This experiment cannot support the abstract's claim of 'improvements in accuracy, especially on small sample sizes'. The same issue appears in attenuated form in Section 4.2, where the ground-truth signatures are drawn from COSMIC and the MVN prior covariance is also estimated from COSMIC.
  2. [§4.2–§4.4] All simulation results are based on one dataset in Section 4.2 and two datasets in Section 4.3, with no multiple seeds, no error bars, and no quantification of MCMC variability. The convergence-iteration differences in Figure 4 and the cosine-similarity differences in Figures 5–7 are therefore point estimates with unknown variability. Since the headline efficiency and accuracy claims rest on these figures, the simulations need to be repeated over many simulated datasets and summarized with appropriate measures of spread.
  3. [§4.1.2] The permutation test does not simulate the null hypothesis it claims to test. Under the null, each mutation type is drawn independently from a univariate truncated normal, but the sampled columns are then normalized to sum to one. Normalization induces negative dependence among the 96 entries, so the simulated null correlation distribution is not one of independence. The result that 319/4560 approximately 6.996% of adjusted p-values are significant therefore does not establish that dependence in the COSMIC signatures exceeds chance; it may simply reflect the normalization artifact.
  4. [§4.4–§4.5] The hierarchical model is validated only on data generated from its own three-parameter covariance and on one real dataset. In the simulation, the recovered rho_same=0.21 and rho_diff=0.08 are not close to the generative values 0.5 and -0.1, so the claim that these are 'close to the original parameter values' is inaccurate, even after accounting for the scale non-identifiability of sigma2_P. On Panc-Endocrine the MAP estimates are near zero (rho_same=0.053, rho_diff=-0.008), which undercuts the motivating assumption for that dataset, and the claimed superiority of the hierarchical model is based only on a visual comparison of cosine-similarity heatmaps without a formal test or uncertainty estimate. This does not provide evidence that the hierarchical model learns dependence structure in real applications.
minor comments (5)
  1. [§4.4 and Figure 7] The text states N=3 and G=60 for the hierarchical simulation, but the Figure 7 caption reports a sample size of 120, and the following paragraph refers to 'a smaller sample size (of 30 or 60 instead of 120)'. This inconsistency should be resolved.
  2. [§3.2.2] There is a duplicated phrase, 'we will still perform the same the same Gibbs updates', which should be corrected.
  3. [General] The manuscript repeatedly refers to itself as 'this thesis' and 'this project' (e.g., Sections 2.1 and 5.1). For a journal submission, this framing should be replaced with neutral language.
  4. [§7.2] The implemented sampler uses truncation bounds at mean plus or minus 10 standard deviations rather than the stated [0, infinity) truncation. This changes the effective prior, and the paper should state this approximation in the main text and discuss its potential impact on the posterior.
  5. [References] Some references have incomplete or irregular author formatting, for example 'Wilhelm and G, 2023' and 'V.' in the Favero et al. entry. These should be brought into consistent journal style.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the multivariate-truncated-normal derivation is self-contained, and the small-sample accuracy comparison is not shown to be an oracle-fed construction.

full rationale

The paper's core derivation—the full conditional for each column of P under a Truncated Multivariate Normal prior—is derived from first principles in Section 7.4, and Section 3.1.2 explicitly shows that substituting the identity matrix for Sigma recovers the nonMVN model. That reduction is internal to the paper and not circular. The accuracy comparison in Section 4.3 generates data from the 3-parameter covariance model, but the paper does not state that the MVN method is given that generative covariance as its prior; the prior covariance source is described in Section 3.1.4 as the James-Stein shrinkage estimate of the COSMIC correlation matrix. Without an explicit statement that the MVN prior equals the simulation's generative Sigma, the 6/8 versus 4/8 result cannot be pinned down as a by-construction artifact. The hierarchical-model simulation in Section 4.4 is a standard generative-model recovery check, and the paper itself describes it as verifying the algorithm's ability to uncover parameters from data generated under its own structure, which is an implementation validation rather than a circular prediction. The main self-citations—Landy et al. for the Normal-likelihood approximation and the bayesNMF implementation—support modeling choices and convergence tools but do not smuggle in the target result that covariance structure improves accuracy. Overall, the derivation chain is self-contained, and the empirical claims, while limited in scope, are not shown to reduce to their own inputs.

Assumptions & free parameters 4 free parameters · 4 assumptions · 0 invented entities

The model leans on external information (COSMIC) for the prior covariance and on a Normal approximation from prior work. The hierarchical model introduces a parsimonious but restrictive covariance parameterization, and the permutation test's null construction is flawed because normalization creates correlation.

free parameters (4)
  • James-Stein shrinkage lambda = 0.0991
    Shrinkage intensity for the prior covariance matrix, estimated from COSMIC signatures (Section 3.1.4).
  • MH proposal standard deviation = 0.1
    Default random walk step size for the hierarchical model's Metropolis-Hastings updates (Section 3.2.2).
  • Truncated MVN bound multiplier = 10
    Lower and upper truncation bounds set to 10 standard deviations from the mean to ensure numerical stability (Section 7.2).
  • Hierarchical hyperparameters = alpha=2, alpha_same=2, alpha_diff=2, beta=2, beta_same=2, beta_diff=2
    All six hyperparameters set to 2; described as vague and uninformative (Section 3.2.1).
assumptions (4)
  • domain assumption Normal likelihood approximates Poisson likelihood for fixed rank NMF.
    Section 1.6: the paper relies on Landy et al. (2025) to justify using Normal likelihood, which is essential for conjugate MVN updates.
  • domain assumption COSMIC signature matrix provides a relevant external covariance prior.
    Section 3.1.4: the prior covariance is the James-Stein shrunk correlation of COSMIC v3.3.1 signatures; transferability to other datasets is assumed.
  • ad hoc to paper The rho_same/rho_diff parameterization captures the dependence structure of mutation types.
    Section 3.2.1: a single shared variance and two correlation parameters are assumed; the authors acknowledge this is a simplification.
  • ad hoc to paper Normalizing independent truncated normal samples preserves independence for the null distribution.
    Section 4.1.2: the permutation test normalizes simulated signatures to sum to one, which introduces correlation; the null of independence is therefore not correctly implemented.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bayesian Non-Negative Matrix Factorization with Correlated Mutation Type Probabilities for Mutational Signatures." pith.science (2026). https://pith.science/paper/JVCL2232

@misc{pith2026250615855,
  author       = {Pith},
  title        = {Pith review of: Bayesian Non-Negative Matrix Factorization with Correlated Mutation Type Probabilities for Mutational Signatures},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JVCL2232}},
  note         = {Machine review of arXiv:2506.15855}
}
read the original abstract

Somatic mutations, or alterations in DNA of a somatic cell, are key markers of cancer. In recent years, mutational signature analysis has become a prominent field of study within cancer research, commonly with Nonnegative Matrix Factorization (NMF) and Bayesian NMF. However, current methods assume independence across mutation types in the signatures matrix. This paper expands upon current Bayesian NMF methodologies by proposing novel methods that account for the dependencies between the mutation types. First, we implement the Bayesian NMF specification with a Multivariate Truncated Normal prior on the signatures matrix in order to model the covariance structure using external information, in our case estimated from the COSMIC signatures database. This model converges in fewer iterations, using MCMC, when compared to a model with independent Truncated Normal priors on elements of the signatures matrix and results in improvements in accuracy, especially on small sample sizes. In addition, we develop a hierarchical model that allows the covariance structure of the signatures matrix to be discovered rather than specified upfront, giving the algorithm more flexibility. This flexibility for the algorithm to learn the dependence structure of the signatures allows a better understanding of biological interactions and how these change across different types of cancer. The code for this project is contributed to an open-source R software package. Our work lays the groundwork for future research to incorporate dependency structure across mutation types in the signatures matrix and is also applicable to any use of NMF beyond just single-base substitution (SBS) mutational signatures.

Figures

Figures reproduced from arXiv: 2506.15855 by the authors.

Figure 1
Figure 1. Heatmap of COSMIC correlations between pairs of the 96 mutation types, sorted by center, left, and right [PITH_FULL_IMAGE:figures/full_fig_p012_1.png] view at source ↗
Figure 2
Figure 2. Distribution of COSMIC correlations for pairs of mutation types with the same and with different center [PITH_FULL_IMAGE:figures/full_fig_p012_2.png] view at source ↗
Figure 3
Figure 3. p-values before and after Benjamini-Hochberg adjustment. [PITH_FULL_IMAGE:figures/full_fig_p013_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Comparison of number of iterations until convergence using MVN and nonMVN methods for simulation [PITH_FULL_IMAGE:figures/full_fig_p014_4.png]
Figure 5
Figure 5. Figure 5: Comparison of cosine similarities between estimated and COSMIC reference signatures using MVN and [PITH_FULL_IMAGE:figures/full_fig_p015_5.png]
Figure 6
Figure 6. Figure 6: Cosine similarity heatmap between estimated (x-axis) and COSMIC reference (y-axis) mutational signatures [PITH_FULL_IMAGE:figures/full_fig_p017_6.png]
Figure 7
Figure 7. Figure 7: Cosine similarity heatmap between estimated (x-axis) and COSMIC reference (y-axis) mutational signatures [PITH_FULL_IMAGE:figures/full_fig_p018_7.png]
Figure 8
Figure 8. Figure 8: Simulation Study Left figure: trace plots of full chain. Right figure: posterior distribution of the 1000 samples immediately before convergence for σ 2 P , ρsame, ρdiff using Hierarchical model for dataset simulated from the 3-parameter structure (σ 2 P = 7, ρsame = 0…
Figure 9
Figure 9. Figure 9: Panc Endocrine data example. Left figure: trace plots of full chain. Right figure: posterior distribution of the 1000 samples immediately before convergence for σ 2 P , ρsame, ρdiff using Hierarchical model for Panc-Endocrine dataset. The final 1000 iterations immediat…
Figure 10
Figure 10. Figure 10: Panc Endocrine data example. Cosine similarity heatmaps between estimated (x-axis) and COSMIC reference (y-axis) mutational signatures for Panc-Endocrine histology group. MVN, nonMVN, and Hierarchical methods with sample size of 77 and rank (number of signatures) of 6…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

34 extracted references · 27 canonical work pages

  1. [10]

    Probabilistic interpretations of non-negative matrix factorization

    Ali Taylan Cemgil. Probabilistic interpretations of non-negative matrix factorization. Computational Intelligence and Neuroscience, 2009:785152,

    work page 2009

  2. [14]

    URL http://dx.doi.org/10.1007/978-3-642-00599-2_68

    doi:10.1007/978-3- 642-00599-2_68. URL http://dx.doi.org/10.1007/978-3-642-00599-2_68 . Hiroyasu Abe and Hiroshi Yadohisa. A non-negative matrix factorization model based on the zero-inflated tweedie distribution. Computational Statistics, 32:475–499,

    doi:10.1007/978-3-

  3. [18]

    doi:10.1093/bioinformatics/btad756

    ISSN 1367-4811. doi:10.1093/bioinformatics/btad756. URL https://doi.org/10.1093/bioinformatics/btad756. S Kasar, J Kim, R Improgo, G Tiao, P Polak, N Haradhvala, MS Lawrence, A Kiezun, SM Fernandes, S Bahl, et al. Whole-genome sequencing reveals activation-induced cytidine deaminase signatures during indolent chronic lymphocytic leukaemia evolution. Natur...

    work page doi:10.1093/bioinformatics/btad756

  4. [20]

    doi:10.1093/bioinformatics/btaa895

    ISSN 1367-4803. doi:10.1093/bioinformatics/btaa895. URL https://doi.org/10.1093/bioinformatics/btaa895. Alessandro Zito and Jeffrey W. Miller. Compressive bayesian non-negative matrix factorization for mutational signatures analysis,

    work page doi:10.1093/bioinformatics/btaa895

  5. [22]

    Yuichi Shiraishi, Georg Tremmel, Satoru Miyano, and Matthew Stephens

    doi:10.3390/cancers16020374. Yuichi Shiraishi, Georg Tremmel, Satoru Miyano, and Matthew Stephens. A simple model-based approach to inferring and visualizing cancer mutation signatures.PLOS Genetics, 11(12):1–21, 12

    work page doi:10.3390/cancers16020374

  6. [24]

    doi:10.1093/bioinformatics/btaa433

    ISSN 1367-4803. doi:10.1093/bioinformatics/btaa433. URL https://doi.org/10.1093/bioinformatics/btaa433. The ICGC/TCGA Pan-Cancer Analysis of Whole Genomes Consortium. Pan-cancer analysis of whole genomes. Nature, 578:82–93,

    work page doi:10.1093/bioinformatics/btaa433

  7. [25]

    doi:10.1038/s41586-020-1969-6. X. Hu, Z. Xu, and S. De. Characteristics of mutational signatures of unknown etiology. NAR Cancer, 2(3):zcaa026, Sep

    work page doi:10.1038/s41586-020-1969-6 1969

  8. [26]

    Ludmil B

    doi:10.1093/narcan/zcaa026. Ludmil B. Alexandrov, Young Seok Ju, Kerstin Haase, Peter Van Loo, Iñigo Martincorena, Serena Nik-Zainal, Yasushi Totoki, Akihiro Fujimoto, Hidewaki Nakagawa, Tatsuhiro Shibata, Peter J. Campbell, Paolo Vineis, David H. Phillips, and Michael R. Stratton. Mutational signatures associated with tobacco smoking in human cancer. Sci...

    work page doi:10.1093/narcan/zcaa026

Show all 34 references
  1. [27]

    Ji-Eun Park, Markia A

    doi:10.1126/science.aag0299. Ji-Eun Park, Markia A. Smith, Sarah C. Van Alsten, Andrea Walens, Di Wu, Katherine A. Hoadley, Melissa A. Troester, and Michael I. Love. Diffsig: Associating risk factors with mutational signatures. bioRxiv,

    doi

  2. [28]

    URL https://www.biorxiv.org/content/early/2023/02/10/2023.02

    doi:10.1101/2023.02.09.527740. URL https://www.biorxiv.org/content/early/2023/02/10/2023.02. 09.527740. F. Maura, A. Degasperi, F. Nadeu, et al. A practical guide for mutational signature analysis in hematological malignancies. Nature Communications, 10:2969,

    2023 doi

  3. [29]

    Christian Heumann

    doi:10.1038/s41467-019-11037-8. Christian Heumann. James–Stein Estimator , pages 699–701. Springer Berlin Heidelberg, Berlin, Heidelberg,

    doi

  4. [31]

    Epub 2005 Nov

    doi:10.2202/1544-6115.1175. Epub 2005 Nov

    2005

  5. [32]

    URL https://doi.org/10.18637/jss.v064

    doi:10.18637/jss.v064.i04. URL https://doi.org/10.18637/jss.v064. i04. Yoav Benjamini and Yosef Hochberg. Controlling the false discovery rate: A practical and powerful approach to multiple testing. Journal of the Royal Statistical Society. Series B (Methodological) , 57(1):289–300,

    doi

  6. [34]

    URL http://www.jstor.org/stable/2986138

    ISSN 00359254, 14679876. URL http://www.jstor.org/stable/2986138. Isabella N. Grabski, Roberta De Vito, Lorenzo Trippa, and Giovanni Parmigiani. Bayesian combinatorial multi-study factor analysis,

  7. [35]

    URL https://arxiv.org/abs/2007.12616. 26 corBayesNMF 7 Supplementary Materials 7.1 Convergence Control Convergence of a Gibbs sampling algorithm can be defined in numerous ways, ranging from monitoring the stabilization of parameter estimates and log-likelihood values to more ...

    2023 arXiv

  8. [1995]

    URL http://www.jstor.org/stable/2346101. W. R. Gilks, N. G. Best, and K. K. C. Tan. Adaptive rejection metropolis sampling within gibbs sampling. Journal of the Royal Statistical Society. Series C (Applied Statistics) , 44(4):455–472,

  9. [1999]

    doi:10.1007/978-1-4757-3071-5_7

    ISBN 978-1-4757-3071-5. doi:10.1007/978-1-4757-3071-5_7. URL https://doi.org/10.1007/ 978-1-4757-3071-5_7 . David M. Blei, Andrew Y . Ng, and Michael I. Jordan. Latent dirichlet allocation. Journal of Machine Learning Research, 3:993–1022,

    doi

  10. [2005]

    URL https://doi.org/10.1016/j.mrfmmm.2004

    doi:10.1016/j.mrfmmm.2004.06.057. URL https://doi.org/10.1016/j.mrfmmm.2004. 06.057. Gerd P. Pfeifer. Environmental exposures and mutational patterns of cancer genomes. Genome Medicine, 2:54,

    2004 doi

  11. [2007]

    Ludmil Alexandrov, Serena Nik-Zainal, David Wedge, et al

    URL http://arxiv.org/abs/0708.4149. Ludmil Alexandrov, Serena Nik-Zainal, David Wedge, et al. Signatures of mutational processes in human cancer. Nature, 500:415–421, 2013b. doi:10.1038/nature12477. URL https://doi.org/10.1038/nature12477. John G Tate, Sally Bamford, Harry C J...

    arXiv

  12. [2009]

    doi:10.1155/2009/785152. Jenna M. Landy, Nishanth Basava, and Giovanni Parmigiani. bayesnmf: Fast bayesian poisson nmf with automatically learned rank applied to mutational signatures,

    2009 doi

  13. [2010]

    URL https://doi.org/10.1186/gm175

    doi:10.1186/gm175. URL https://doi.org/10.1186/gm175. Daniel D. Lee and H. Sebastian Seung. Algorithms for non-negative matrix factorization. In Advances in Neural Information Processing Systems, volume 13, pages 556–562. MIT Press,

    doi

  14. [2011]

    doi:10.1007/978-3-642-04898-2_318

    ISBN 978-3-642-04898-2. doi:10.1007/978-3-642-04898-2_318. URL https://doi.org/10.1007/ 978-3-642-04898-2_318 . Olivier Ledoit and Michael Wolf. Improved estimation of the covariance matrix of stock returns with an application to portfolio selection. Journal of Empirical Finan...

    doi

  15. [2012]

    URL https://doi.org/10.1016/ j.cell.2012.04.024

    doi:10.1016/j.cell.2012.04.024. URL https://doi.org/10.1016/ j.cell.2012.04.024. Arne Van Hoeck, Niels H. Tjoonk, Ruben van Boxtel, et al. Portrait of a cancer: Mutational signature analyses for cancer diagnostics. BMC Cancer, 19:457,

    2012 doi

  16. [2013]

    URL https://doi.org/ 10.1038/ng.2764

    doi:10.1038/ng.2764. URL https://doi.org/ 10.1038/ng.2764. Gerd P. Pfeifer, Young-Hyun You, and Ahmad Besaratinia. Mutations induced by ultraviolet light. Mutation Research, 571(1-2):19–31,

    doi

  17. [2014]

    doi:10.1038/nrg3729

    ISSN 1471-0064. doi:10.1038/nrg3729. URL https://www.nature.com/articles/nrg3729. Publisher: Nature Publishing Group. Serena Nik-Zainal, Ludmil B. Alexandrov, David C. Wedge, et al. Mutational processes molding the genomes of 21 breast cancers. Cell, 149(5):994–1007,

    doi

  18. [2015]

    URL https://doi.org/10.1371/journal.pgen.1005657

    doi:10.1371/journal.pgen.1005657. URL https://doi.org/10.1371/journal.pgen.1005657. Yang Zhang, Yunxuan Xiao, Muyu Yang, and Jian Ma. Cancer mutational signatures representation by large-scale context embedding. Bioinformatics, 36(Supplement_1):i309–i316, 07

    doi

  19. [2016]

    URL http://dx.doi.org/10.1093/bioinformatics/btw572

    doi:10.1093/bioinformatics/btw572. URL http://dx.doi.org/10.1093/bioinformatics/btw572. Shixiang Wang, Ziyu Tao, Tao Wu, and Xue-Song Liu. Sigflow: an automated and comprehensive pipeline for cancer genome mutational signature analysis. Bioinformatics, 37(11):1590–1592, 10

    doi

  20. [2017]

    URL https://doi

    doi:10.1007/s00180-016-0689-8. URL https://doi. org/10.1007/s00180-016-0689-8 . Luiz Paulo Fávero, Jr. Hair, Joseph F., Rafael de Freitas Souza, Matheus Albergaria, and Talles V . Brugni. Zero-inflated generalized linear mixed models: A better way to understand data relationsh...

    doi

  21. [2019]

    URL https://doi.org/10

    doi:10.1186/s12885-019-5677-2. URL https://doi.org/10. 1186/s12885-019-5677-2 . Ludmil B. Alexandrov, Serena Nik-Zainal, David C. Wedge, Peter J. Campbell, and Michael R. Stratton. De- ciphering signatures of mutational processes operative in human cancer. Cell Reports , 3(1):...

    doi

  22. [2020]

    URL https://doi.org/10.1038/ s41586-020-1943-3

    doi:10.1038/s41586-020-1943-3. URL https://doi.org/10.1038/ s41586-020-1943-3 . The Cancer Genome Atlas Research Network, J. N. Weinstein, E. A. Collisson, et al. The cancer genome atlas pan-cancer analysis project. Nature Genetics, 45:1113–1120,

    1943 doi

  23. [2021]

    24 corBayesNMF Christian P

    doi:10.3390/math9101100. 24 corBayesNMF Christian P. Robert and George Casella. The Gibbs Sampler , pages 285–361. Springer New York, New York, NY ,

    doi

  24. [2023]

    Mikkel N

    doi:10.1186/s12859- 023-05304-1. Mikkel N. Schmidt, Ole Winther, and Lars Kai Hansen. Bayesian non-negative matrix factorization. In Independent Component Analysis and Signal Separation , pages 540–547. Springer Berlin Heidelberg,

    doi

  25. [2024]

    Alexandra Dananberg, Josefine Striepen, Jacob S

    URL https://arxiv.org/abs/2404.10974. Alexandra Dananberg, Josefine Striepen, Jacob S. Rozowsky, and Mia Petljak. Apobec mutagenesis in cancer development and susceptibility. Cancers (Basel), 16(2):374, Jan

    arXiv

  26. [2025]

    Abraham Berman and Robert J

    URL https://arxiv.org/abs/2502.18674. Abraham Berman and Robert J. Plemmons. Nonnegative matrices in the mathematical sciences . SIAM, Philadelphia, PA,

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.