Pith. sign in

REVIEW 3 major objections 4 minor 46 references

Multiomics Tissue Segmentation via Spatially-Informed Nested Biclustering Methods

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read The paper claims that multiomics imaging data, in which several molecular classes share one pixel grid, can be modeled as a single nested biclustering problem whose shared tissue segmentation reveals regions and markers that single-omics an

desk verdict Poseidon is a genuine extension of nested Bayesian biclustering to multiomics MSI, with clean derivations and a real simulation win, but the shared-partition assumption is untested and the real-data evidence is exploratory. read the letter →

arxiv 2509.02482 v1 pith:Q2UJVIME submitted 2025-09-02 stat.ME

Francesco Denti , Cecilia Balocchi , Vanna Denti , Giulia Capitoli This is my paper · ORCID
classification stat.ME MSC 62F1562H30
keywords multiomicsMALDI-MSItissuesegmentationnestedbiclusteringBayesiannonparametricsseparateexchangeabilityPottsmodelvariationalinferenceclearcellrenalcarcinoma
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

The paper introduces Poseidon, a Bayesian nonparametric method that jointly segments a tissue section and clusters molecular signals from different omics layers in a single analysis. It targets the setting, newly enabled by a multiomics mass-spectrometry workflow, where lipids, N-glycans, and peptides are measured on the same tissue section over the same pixel grid. The method's claim is that one shared tissue segmentation, informed by all three molecular classes, yields a more coherent picture of tumor heterogeneity than any single class alone, while nested clustering within each region points to the molecules that define it. The authors support the claim with simulations, comparisons against existing biclustering and segmentation approaches, and an application to a clear cell renal cell carcinoma sample in which the model separates tumor grades, hemorrhagic areas, and healthy cortex. The practical stakes are that scarce tissue, such as rare tumors or small biopsies, can yield an integrated spatial molecular map from a single section.

What carries the argument

The load-bearing object is the nested biclustering partition: a shared column partition C over pixels (tissue regions) and, nested inside each region, a dataset-specific row partition over m/z signals, with shared atoms (one pool of cluster parameters all regions draw from) as the cluster parameters. Spatial dependence enters through a BNP-MRF prior, a stick-breaking Dirichlet process modified by a Potts term whose inverse temperature beta rewards neighboring pixels sharing a label. The finite-infinite shared-atoms construction, with sparse Dirichlet row weights and stick-breaking column weights, yields closed-form prior co-clustering probabilities, per Proposition 2.1, lower than those of c

What would settle it

Generate a synthetic multiomics dataset in which the true pixel partition differs across molecular classes, for example lipids have an extra spatial boundary that peptides lack, fit Poseidon, and compare each class's estimated segmentation against that class's own ground truth using the adjusted Rand index; the shared-partition model predicts measurable degradation for the class with unique boundaries. On the biological side, check the top region-defining markers, such as the presumed histone H2A peptide at m/z 944.53, by targeted MS/MS or immunohistochemistry to see whether they localize to t

Watch

Extended reading notes

Core claim

Poseidon models multiomics imaging data as one nested biclustering problem: a shared partition of pixels into tissue regions and, inside each region, a separate partition of that class's m/z signals. The single-dataset building block Pose keeps rows meaningfully labeled, so pixels group together when the same signals are active rather than merely the same fraction of signals. A Potts-type Markov random field prior enforces spatial coherence, and a finite-infinite shared-atoms construction keeps prior co-clustering low. Poseidon keeps one pixel partition across molecular classes while giving each class its own row clusters and parameters. On the kidney tumor data it separates tumor grades, a

Load-bearing premise

The model forces one pixel partition to be shared by all molecular classes; if lipids, N-glycans, and peptides genuinely have different spatial boundaries, that single shared partition will blur or bias the estimated tissue regions.

Editorial extensions

If this is right

  • Multiomics integration gives a more coherent tissue segmentation than any single molecular class or a row-stacked dataset, because complementary signals suppress noise and class-specific artifacts, as demonstrated on the ccRCC case and in simulations.
  • In the multi-dataset simulation, one informative dataset anchors the shared pixel partition when the partner dataset is nearly pure noise, and the noisy dataset's own row clusters are recovered substantially better than in single-dataset analysis, so information pooling works across heterogeneous data quality.
  • The model decouples the shared spatial segmentation from class-specific signal clustering: a region can be defined by the same pixels while each molecular class assigns a different set of m/z signals to it, which is how the analysis distinguishes hemorrhagic, mixed-grade tumor, and healthy cortex.
  • Poseidon's output is directly usable as a low-dimensional summary for pathologists: region maps plus, per region, posterior mean abundance of every m/z colored by row cluster, yielding candidate region-defining molecules.
  • The variational algorithm makes the fit practical at scale, roughly 44 seconds per run for the three-class ccRCC dataset with over 4,500 pixels per class, so the method is feasible in real MSI workflows.

Reading between the lines

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

  • The shared-partition assumption is the paper's weakest commitment: if a molecular class has spatial boundaries the others lack, forcing one common pixel partition will blur or bias that class's segmentation. A direct test would simulate classes with deliberately different ground-truth boundaries and measure each class's segmentation against its own truth.
  • The region-specific molecules the model highlights, such as the presumed histone H2A peptide at m/z 944.53 and the putative Heterochromatin Protein 1-binding protein 3 peptide at m/z 1116.56, are hypotheses rather than validations; independent confirmation via targeted mass spectrometry or immunohistochemistry would test whether the shared-partition constraint created them.
  • The model's structure transfers to other multiomics assays sharing one spatial lattice, such as spatial transcriptomics or combined proteomics-metabolomics acquisitions, if the Gaussian likelihood and quantile transform are replaced with count-appropriate likelihoods.
  • A single inverse temperature beta for all molecular classes may be too coarse; allowing per-class spatial smoothing is a testable extension that could better reflect molecules with different spatial coherence.
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

3 major / 4 minor

Summary. Poseidon is a Bayesian nonparametric nested biclustering model for multiomics MALDI-MSI data. It models several molecular-class matrices observed on a common pixel grid, assuming one shared spatial partition of pixels (induced by a Potts-type BNP-MRF prior) and dataset-specific row partitions with shared atoms. The paper derives prior coclustering probabilities for the proposed shared-atoms formulation, implements mean-field variational inference, evaluates the method on synthetic data against sparseBC and double k-means, and applies it to a ccRCC tissue section with lipid, N-glycan, and peptide measurements.

Significance. If the central modeling assumption is justified, Poseidon provides a principled integrated framework for tissue segmentation and molecular feature clustering in multiomics MSI data. The analytic result in Proposition 2.1 is a useful and clean distinction between shared-atoms and common-atoms priors, and the simulation study in Section 4.1 shows clear gains over sparseBC and double k-means under a common underlying partition. The variational algorithm also appears practical: it handles datasets with roughly 100 features and over 4,500 pixels in about 44 seconds. However, the validation is incomplete in three load-bearing respects: the shared-column-partition assumption is not stress-tested against datasets with genuinely different spatial structures; the variational update for the stick-breaking weights is made under a beta=0 approximation while beta is elsewhere estimated; and the real-data claims are not quantitatively validated against available pathologist annotations. These points must be addressed before the method's central claims can be accepted.

major comments (3)
  1. [Section 2.2, Eq. (3)] The model assumes one column partition C shared by all T datasets. This is the central integrative assumption, but it is only tested in settings where all datasets share the same true partition. In Section 4.2 and S.3.2, every D(r) is generated from the same biclustering structure, so the information-borrowing result is conditional on the exact assumption at issue. The real data in Figure 5 suggest the assumption may be strong: single-molecule Pose fits show lipid-specific intra-tumor clusters that do not appear for N-glycans or peptides. Add simulations with partially or fully divergent true spatial partitions (e.g., an extra cluster in one modality, or shifted boundaries), report per-modality ARI for Poseidon versus separate Pose fits, and discuss possible diagnostics or extensions to partially shared partitions. Without this, the claimed advantage of integration may be an artifact of
  2. [Section 5 and Figure 5] The variational update for v_k is described as 'mimic[king] Lü et al. [2020] and assum[ing] beta=0', while beta itself is estimated via the pseudo-likelihood/A2 approximation and enters the update for rho_{j,k}. This is an internal inconsistency in the variational approximation. No sensitivity analysis is reported: the authors do not compare results for beta fixed at 0, beta estimated on a coarse grid, or beta fixed at values in the prior range, and they do not report the posterior mean or posterior distribution of beta for the real application. Since the spatial Potts term is a central component of Poseidon, the effect of this approximation on the recovered segmentation should be quantified.
  3. [Section 5] The real-data evaluation is entirely qualitative. The manuscript claims that Poseidon yields 'a more coherent and biologically meaningful tissue segmentation' and that it detects regions missed by histopathological examination, but no quantitative comparison is made to the expert pathologist annotations shown in Figure S.1, even though those ROIs are available. I recommend computing overlap measures (ARI or Dice) between the Poseidon segmentation and the annotated regions, and comparing these against separate Pose fits and the stacked-data fit. At minimum, report the stability of the segmentation across random initializations and the posterior uncertainty of the relevant cluster assignments. This is necessary to support the applied claims in Section 5.
minor comments (4)
  1. [Algorithm 1, Step 2] The variational update defines q*(r^{(t)}_{i,k}=l) = xi^{(t)}_{i,j,l}; the subscript j is erroneous and should be k (the density does not depend on a specific pixel j).
  2. [Section S.4] The text says 'in Figure S.4, we illustrate the spatial distribution' but the referenced figure is Figure S.9. Please correct the cross-reference.
  3. [Equation (4)] The min-max standardization followed by the probit transform is a substantial preprocessing choice. Its effect on variance, outliers, and between-class comparability is not discussed or explored. A sentence justifying this transformation and perhaps a robustness check would help.
  4. [General] No code or data availability statement is provided. For reproducibility of the simulation and real-data analyses, please include an availability statement or clearly state that code/data are available upon request.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Poseidon's model, inference, and evaluation are specified in-paper; self-citations are background, not load-bearing reductions.

full rationale

The paper's derivation chain is self-contained. Section 2 defines the single-dataset Pose model and the multi-dataset Poseidon model explicitly (Eqs. 1-3), including priors, likelihoods, and the shared column partition assumption. Proposition 2.1 is a mathematical derivation from those prior definitions, not an empirical prediction or fitted constant. The variational inference algorithm in Section 3 is derived from the model's full conditionals, with approximations clearly stated (e.g., pseudo-likelihood for the Potts term). Simulations in Section 4 compare Pose/Poseidon against external baselines (sparseBC, Double k-means) on synthetic data with known ground truth; no parameter is fitted to the target outcome and then renamed as a prediction. Real-data results in Section 5 are posterior summaries and cluster visualizations, not out-of-sample predictions forced by fitted inputs. The shared column partition is a modeling assumption that may be biologically misspecified, but it is an input assumption, not a conclusion derived from itself. Self-citations (Denti et al. 2022 for the MSI workflow; Denti et al. 2023 and D'Angelo and Denti 2024 for nested priors) are external methodological building blocks; the model equations are re-specified in the paper and do not reduce to those citations. Section 6 acknowledges limitations (e.g., Potts computational challenges, alternative priors), further indicating that the claims are not presented as forced by a self-citation chain.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central claim rests on a small number of hand-set hyperparameters and on two domain assumptions that are stated but not tested: the shared column partition across molecular classes, and the Gaussian mixture after a probit transform. The Potts inference also relies on an approximation whose error is unquantified.

free parameters (4)
  • b0 (Dirichlet concentration for row mixture weights) = 0.0001 * 1_L
    Set by hand in Section 2.2 to force sparse finite mixtures over rows; controls the number and size of row clusters.
  • L and K (truncation bounds for row and column clusters) = L=40, K=30 in application; L=10, K=10 or L=30, K=15 in simulations
    Chosen by the analyst; they cap the support of the nonparametric priors and affect the inferred segmentation and cluster counts.
  • beta (inverse temperature of the Potts prior) = beta=1 in Section 4.1, beta=0 in Section 4.2, U(0,2) prior in Section 5
    Controls spatial smoothing strength; in the application it is inferred only approximately through a pseudo-likelihood approximation, and in simulations it is fixed by hand.
  • Prior hyperparameters (m0, k0, c0, d0) = (0, 0.1, 3, 2)
    Recommended values for the Normal-Inverse-Gamma base measures, chosen by hand without sensitivity analysis.
assumptions (5)
  • domain assumption Separate exchangeability: each row is the same m/z feature across all pixels, so row allocations are nested within column clusters.
    Introduced in Section 2.1 to preserve row identity; if rows do not have consistent meaning across pixels, the nested structure is not justified.
  • domain assumption A single shared column partition exists across all molecular datasets.
    Section 2.2, model (3): 'we assume the existence of one partition of the columns C ... shared by all the datasets'. This is the core integration assumption and is not tested.
  • domain assumption The abundance matrix entries follow a Gaussian mixture after the probit-style transform (4).
    Section 2.2 uses f as a Gaussian density after min-max scaling followed by inverse normal cdf; the transformation is ad hoc and no adequacy check is reported.
  • ad hoc to paper The Potts normalizing constant can be approximated by pseudo-likelihood and swapping expectation with log (approximations PL and A2 in Section S.2.2).
    This approximation is introduced to make variational inference tractable; its accuracy is not assessed.
  • standard math Standard stochastic process machinery: GEM stick-breaking, Dirichlet distributions, Normal-Inverse-Gamma conjugacy, and mean-field factorization.
    Used throughout Sections 2 and 3; unproved background tools.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Multiomics Tissue Segmentation via Spatially-Informed Nested Biclustering Methods." pith.science (2026). https://pith.science/paper/Q2UJVIME

@misc{pith2026250902482,
  author       = {Pith},
  title        = {Pith review of: Multiomics Tissue Segmentation via Spatially-Informed Nested Biclustering Methods},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q2UJVIME}},
  note         = {Machine review of arXiv:2509.02482}
}
read the original abstract

Matrix-Assisted Laser Desorption/Ionisation Mass Spectrometry Imaging (MSI) is a powerful technique for spatially resolved molecular profiling and cancer biomarker discovery. Recent advances, including a novel multiomics workflow, enable multiple rounds of MSI on the same tissue section, extracting diverse molecular classes, e.g., lipids, peptides, and N-glycans, while preserving spatial resolution. This innovation is particularly valuable for studies with limited tissue sample availability, such as rare diseases or small tumors. However, the resulting data are high-dimensional, spatially structured, and the various molecular types share the same pixel grid. To address these challenges, we propose Poseidon, a Bayesian nonparametric nested biclustering model that simultaneously segments the common tissue pixels and clusters molecular signals within each molecular class, leveraging the shared spatial structure. A separately exchangeable framework is first considered, and then extended to handle spatial data via hidden Markov random fields. For scalability, we implement an efficient mean-field variational inference algorithm tailored for multi-dataset analysis. After validating the efficacy of our method on simulated scenarios, we demonstrate the applicability of our model in a real-world case study, where multiomics measurements were performed on kidney tissue affected by clear cell renal cell carcinoma. The nested, hierarchical structure of Poseidon, combined with its principled inferential framework, allows the extraction of interesting biological insights, such as clear tissue segmentation and biomarker detection.

Figures

Figures reproduced from arXiv: 2509.02482 by the authors.

Figure 1
Figure 1. A depiction of the multiomics workflow developed in Denti et al. [2022]. The same tissue undergoes multiple consecutive MSI analyses with MALDI, yielding different MSI datasets – one for each molecular class. for the analysis of multiomics MSI data based on nested biclustering, which simultaneously segments spatial pixels and clusters molecular signals within each tissue subregion. Our approach integrates the multip… view at source ↗
Figure 2
Figure 2. The top row displays the Hematoxylin and Eosin stained image of the ccRCC tissue used in the analysis (A) and heatmaps displaying the (normalized) median abundance across the tissue sample for lipids, N-glycans, and peptides (B-D). The bottom row reports, for a specific highlighted pixel (E), abundances of the signals for the different molecular classes (F-H). 2 A finite-infinite separately exchangeable model for MA… view at source ↗
Figure 3
Figure 3. A diagram summarizing the partition structure we aim to recover. The partition over the columns is shared across datasets. The partition over the rows is dataset-specific. this translates into simultaneously clustering the columns and the matrix entries. Most of these nested models were proposed under the hypothesis of partial exchangeability, where data are assumed to be exchangeable within the different subpopulat… view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Recovery of column partition (left panel) and biclustering structure (central and right panel), measured by ARI (left and center) and RMSE (right) in synthetic data. We consider five datasets, denoted D(r) , each with J = 20 columns and N(r) = 50 rows, for r = 0, . . .…
Figure 5
Figure 5. Figure 5: Tissue segmentation obtained with fitting single-dataset Pose models to the separate molecules (top panels, A–C) and the stacked datasets (bottom left panel, D). These results are compared with Poseidon’s multi-dataset approach (bottom right panel, E). areas with hemor…
Figure 6
Figure 6. Figure 6: Spatial and spectral characterization of three representative regions of interest identified within the tissue. The left panels show the spatial distribution of each CC, highlighting the localization patterns across the sample. Right panels dis￾play the corresponding p…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

46 extracted references · 41 canonical work pages

  1. [1]

    Efficient spatial segmentation of large imaging mass spectrometry datasets with spatially aware clustering

    Theodore Alexandrov and Jan Hendrik Kobarg. Efficient spatial segmentation of large imaging mass spectrometry datasets with spatially aware clustering. Bioinformatics, 27 0 (13): 0 i230–i238, June 2011. ISSN 1367-4803. doi:10.1093/bioinformatics/btr246. URL http://dx.doi.org/10.1093/bioinformatics/btr246

    work page doi:10.1093/bioinformatics/btr246 2011

  2. [2]

    Integrative clustering in mass spectrometry imaging for enhanced patient stratification

    Benjamin Balluff, Achim Buck, Marta Martin-Lorenzo, Fr \'e d \'e ric Dewez, Rupert Langer, Liam A McDonnell, Axel Walch, and Ron MA Heeren. Integrative clustering in mass spectrometry imaging for enhanced patient stratification. PROTEOMICS--Clinical Applications, 13 0 (1): 0 1800137, 2019

    work page 2019

  3. [3]

    Clustering Areal Units at Multiple Levels of Resolution to Model Crime in Philadelphia

    Cecilia Balocchi, Edward I. George, and Shane T. Jensen. Clustering Areal Units at Multiple Levels of Resolution to Model Crime in Philadelphia . ArXiv preprint , 2021. URL http://arxiv.org/abs/2112.02059

    work page Pith review arXiv 2021

  4. [4]

    Quintana

    Mario Beraha, Alessandra Guglielmi, and Fernando A. Quintana. The Semi-Hierarchical Dirichlet Process and Its Application to Clustering Homogeneous Distributions . Bayesian Analysis, 16 0 (4): 0 1187--1219, 2021. ISSN 19316690. doi:10.1214/21-BA1278

    work page doi:10.1214/21-ba1278 2021

  5. [5]

    Spatial interaction and the statistical analysis of lattice systems

    Julian Besag. Spatial interaction and the statistical analysis of lattice systems. Journal of the Royal Statistical Society: Series B, 36 0 (2): 0 192--236, 1974

    work page 1974

  6. [6]

    On the Statistical Analysis of Dirty Pictures

    Julian Besag. On the Statistical Analysis of Dirty Pictures . Journal of the Royal Statistical Society. Series B (Methodological), 48 0 (3): 0 259--302, 1986

    work page 1986

  7. [7]

    Lipid metabolism and cancer

    Xueli Bian, Rui Liu, Ying Meng, Dongming Xing, Daqian Xu, and Zhimin Lu. Lipid metabolism and cancer. Journal of Experimental Medicine, 218 0 (1): 0 1--34, 2021

    work page 2021

  8. [8]

    Blei, Alp Kucukelbir, and Jon D

    David M. Blei, Alp Kucukelbir, and Jon D. McAuliffe. Variational Inference: A Review for Statisticians . Journal of the American Statistical Association, 112 0 (518): 0 859--877, 2017. doi:10.1080/01621459.2017.1285773

    arXiv 2017

Show all 46 references
  1. [9]

    Dunson, Antonio Lijoi, Igor Pr \" u nster, and Abel Rodr \' i guez

    Federico Camerlenghi, David B. Dunson, Antonio Lijoi, Igor Pr \" u nster, and Abel Rodr \' i guez. Latent Nested Nonparametric Priors (with Discussion) . Bayesian Analysis, 14 0 (4): 0 1--54 , 2019. doi:10.1214/19-ba1169. URL http://arxiv.org/abs/1801.05048

    2019 arXiv

  2. [10]

    Giulia Capitoli, Kirsten C. J. van Abeelen, Isabella Piga, Vincenzo L’Imperio, Marco S. Nobile, Daniela Besozzi, and Stefania Galimberti. Well Begun Is Half Done: The Impact of Pre-Processing in MALDI Mass Spectrometry Imaging Analysis Applied to a Case Study of Thyroid Nodule...

    2025

  3. [11]

    A finite-infinite shared atoms nested model for the bayesian analysis of large grouped data

    Laura D'Angelo and Francesco Denti. A finite-infinite shared atoms nested model for the bayesian analysis of large grouped data. Bayesian Analysis, In press: 0 1--34, 2024

    2024

  4. [12]

    A Common Atoms Model for the Bayesian Nonparametric Analysis of Nested Data

    Francesco Denti, Federico Camerlenghi, Michele Guindani, and Antonietta Mira. A Common Atoms Model for the Bayesian Nonparametric Analysis of Nested Data . Journal of the American Statistical Association, 118 0 (541): 0 405--416, 2023

    2023

  5. [13]

    Spatial multiomics of lipids, n-glycans, and tryptic peptides on a single ffpe tissue section

    Vanna Denti, Giulia Capitoli, Isabella Piga, Francesca Clerici, Lisa Pagani, Lucrezia Criscuolo, Greta Bindi, Lucrezia Principi, Clizia Chinello, Giuseppe Paglia, et al. Spatial multiomics of lipids, n-glycans, and tryptic peptides on a single ffpe tissue section. Journal of P...

    2022

  6. [14]

    Vanna Denti, Angela Greco, Antonio Maria Alviano, Giulia Capitoli, Nicole Monza, Andrew Smith, Daniela Pilla, Alice Maggioni, Mariia Ivanova, Konstantinos Venetis, et al. Spatially resolved molecular characterization of noninvasive follicular thyroid neoplasms with papillary-l...

    2024

  7. [15]

    J. B. Durand, F. Forbes, C. D. Phan, L. Truong, H. D. Nguyen, and F. Dama. Bayesian non-parametric spatial prior for traffic crash risk mapping: A case study of Victoria, Australia . Australian and New Zealand Journal of Statistics, 64 0 (2): 0 171--204, 2022

    2022

  8. [16]

    Nelson, Jiahui Li, Chao Wu, Jimeng Yang, Yan Zheng, Christiane Bruns, Yue Zhao, Lunxiu Qin, and Qiongzhu Dong

    Yan Fu, Tiantian Zou, Xiaotian Shen, Peter J. Nelson, Jiahui Li, Chao Wu, Jimeng Yang, Yan Zheng, Christiane Bruns, Yue Zhao, Lunxiu Qin, and Qiongzhu Dong. Lipid metabolism in cancer progression and therapeutic strategies. MedComm, 2: 0 27--59, 2021. ISSN 2688-2663. doi:10.10...

    2021 doi

  9. [17]

    Proteomics and its current application in biomedical area: Concise review

    Semira Gobena, Bemrew Admassu, Mebrie Zemene Kinde, and Abebe Tesfaye Gessese. Proteomics and its current application in biomedical area: Concise review. The Scientific World Journal, 2024 0 (1): 0 4454744, 2024

    2024

  10. [18]

    Co-clustering: models, algorithms and applications

    G \'e rard Govaert and Mohamed Nadif. Co-clustering: models, algorithms and applications. John Wiley & Sons, 2013

    2013

  11. [19]

    Unsupervised segmentation of mass spectrometric ion images characterizes morphology of tissues

    Dan Guo, Kylie Bemis, Catherine Rawlins, Jeffrey Agar, and Olga Vitek. Unsupervised segmentation of mass spectrometric ion images characterizes morphology of tissues. Bioinformatics, 35 0 (14): 0 i208--i217, 2019

    2019

  12. [20]

    Renal cell carcinoma

    James J Hsieh, Mark P Purdue, Sabina Signoretti, Charles Swanton, Laurence Albiges, Manuela Schmidinger, Daniel Y Heng, James Larkin, and Vincenzo Ficarra. Renal cell carcinoma. Nature reviews Disease primers, 3 0 (1): 0 1--19, 2017

    2017

  13. [21]

    Genomic classifications of renal cell carcinoma: A critical step towards the future application of personalized kidney cancer care with pan-omics precision

    James J Hsieh, Valerie Le, Dengfeng Cao, Emily H Cheng, and Chad J Creighton. Genomic classifications of renal cell carcinoma: A critical step towards the future application of personalized kidney cancer care with pan-omics precision. The Journal of pathology, 244 0 (5): 0 525...

    2018

  14. [22]

    State-of-the-art mass spectrometry imaging applications in biomedical research

    Kasper K Krestensen, Ron MA Heeren, and Benjamin Balluff. State-of-the-art mass spectrometry imaging applications in biomedical research. Analyst, 148 0 (24): 0 6161--6187, 2023

    2023

  15. [23]

    A nonparametric Bayesian model for local clustering with application to proteomics

    Juhee Lee, Peter M \" u ller, Yitan Zhu, and Yuan Ji. A nonparametric Bayesian model for local clustering with application to proteomics . Journal of the American Statistical Association, 108 0 (503): 0 775--788, 2013. ISSN 01621459. doi:10.1080/01621459.2013.784705

    2013

  16. [24]

    Flexible clustering via hidden hierarchical Dirichlet priors

    Antonio Lijoi, Igor Pr \" u nster, and Giovanni Rebaudo. Flexible clustering via hidden hierarchical Dirichlet priors . Scandinavian Journal of Statistics, 50 0 (1): 0 213--234, 2023. ISSN 14679469. doi:10.1111/sjos.12578

    2023 doi

  17. [25]

    Linderman, Matthew J

    Scott W. Linderman, Matthew J. Johnson, and Ryan P. Adams. Dependent multinomial models made easy: Stick‑breaking with the pólya‑gamma augmentation. In Advances in Neural Information Processing Systems 28, pages 3456--3464, 2015

    2015

  18. [26]

    Bayesian nonparametric priors for hidden Markov random fields

    Hongliang L \" u , Julyan Arbel, and Florence Forbes. Bayesian nonparametric priors for hidden Markov random fields . Statistics and Computing, 30 0 (4): 0 1015--1035, 2020

    2020

  19. [27]

    cluster: Cluster Analysis Basics and Extensions, 2025

    Martin Maechler, Peter Rousseeuw, Anja Struyf, Mia Hubert, and Kurt Hornik. cluster: Cluster Analysis Basics and Extensions, 2025

    2025

  20. [28]

    u hwirth-Schnatter, and Bettina Gr \

    Gertraud Malsiner-Walli, Sylvia Fr \"u hwirth-Schnatter, and Bettina Gr \"u n. Model-based Clustering Based on Sparse Finite Gaussian Mixtures . Statistics and Computing, 26: 0 303--324, 2016. doi:10.1007/s11222-014-9500-2

    2016 doi

  21. [29]

    C. A. McGrory, D. M. Titterington, R. Reeves, and A. N. Pettitt. Variational bayes for estimating the parameters of a hidden potts model . Statistics and Computing, 19 0 (3): 0 329--340, 2009

    2009

  22. [30]

    Prospective on imaging mass spectrometry in clinical diagnostics

    Jessica L Moore, Nathan Heath Patterson, Jeremy L Norris, and Richard M Caprioli. Prospective on imaging mass spectrometry in clinical diagnostics. Molecular & Cellular Proteomics, 22 0 (9): 0 1--10, 2023

    2023

  23. [31]

    Nicholls, Anthony N

    Matthew Moores, Geoff K. Nicholls, Anthony N. Pettitt, and Kerrie Mengersen. Scalable bayesian inference for the inverse temperature of a hidden potts model . Bayesian Analysis, 15 0 (1): 0 1--27, 2020

    2020

  24. [32]

    Iain Murray, Zoubin Ghahramani, and David J.C. MacKay. MCMC for doubly-intractable distributions . Proceedings of the 22nd Conference on Uncertainty in Artificial Intelligence, UAI 2006, pages 359--366, 2006

    2006

  25. [33]

    Evaluation and comparison of unsupervised methods for the extraction of spatial patterns from mass spectrometry imaging data (msi)

    Mridula Prasad, Geert Postma, Pietro Franceschi, Lutgarde MC Buydens, and Jeroen J Jansen. Evaluation and comparison of unsupervised methods for the extraction of spatial patterns from mass spectrometry imaging data (msi). Scientific Reports, 12 0 (1): 0 15687, 2022

    2022

  26. [34]

    Separate Exchangeability as Modeling Principle in Bayesian Nonparametrics

    Giovanni Rebaudo, Qiaohui Lin, and Peter Mueller. Separate Exchangeability as Modeling Principle in Bayesian Nonparametrics . ArXiv Preprint, 2021

    2021

  27. [35]

    Dunson, and Alan E

    Abel Rodr \' i guez, David B. Dunson, and Alan E. Gelfand. The nested dirichlet process . Journal of the American Statistical Association, 103 0 (483): 0 1131--1154, 2008. ISSN 01621459. doi:10.1198/016214508000000553

    2008 doi

  28. [36]

    Rohner, D

    T.C. Rohner, D. Staab, and M. Stoeckli. Maldi mass spectrometric imaging of biological tissue sections. Mechanisms of Ageing and Development, 126 0 (1): 0 177--185, 2005

    2005

  29. [37]

    Ortori, David A

    Mohamad Nasir Shafiee, Catharine A. Ortori, David A. Barrett, Nigel P. Mongan, Jafaru Abu, and William Atiomo. Lipidomic biomarkers in polycystic ovary syndrome and endometrial cancer. International Journal of Molecular Sciences, 21: 0 1--15, 2020. ISSN 14220067. doi:10.3390/i...

    2020 doi

  30. [38]

    Moremen, Nathan E

    Pamela Stanley, Kelley W. Moremen, Nathan E. Lewis, Naoyuki Taniguchi, and Markus Aebi. N-Glycans. Cold Spring Harbor Laboratory Press, Cold Spring Harbor (NY), 4th edition, 2022. ISBN 9781621824220. URL http://europepmc.org/books/NBK579964

    2022

  31. [39]

    Sparse biclustering of transposable data

    Kean Ming Tan and Daniela M Witten. Sparse biclustering of transposable data. Journal of Computational and Graphical Statistics, 23 0 (4): 0 985--1008, 2014

    2014

  32. [40]

    Estimating the number of clusters in a data set via the gap statistic

    Robert Tibshirani, Guenther Walther, and Trevor Hastie. Estimating the number of clusters in a data set via the gap statistic. Journal of the Royal Statistical Society: Series B , 63 0 (2): 0 411--423, 2001

    2001

  33. [41]

    Maldi-msi towards multimodal imaging: Challenges and perspectives

    Michael Tuck, Florent Grélard, Landry Blanc, and Nicolas Desbenoit. Maldi-msi towards multimodal imaging: Challenges and perspectives. Frontiers in Chemistry, 10: 0 1--11, 2022. ISSN 2296-2646. doi:10.3389/fchem.2022.904688. URL http://dx.doi.org/10.3389/fchem.2022.904688

    2022

  34. [42]

    Lipids in cancer: a global view of the contribution of lipid pathways to metastatic formation and treatment resistance

    Sophie Vasseur and Fabienne Guillaumond. Lipids in cancer: a global view of the contribution of lipid pathways to metastatic formation and treatment resistance. Oncogenesis, 11 0 (1): 0 46, 2022

    2022

  35. [43]

    Caprioli, and Raf Van de Plas

    Nico Verbeeck, Richard M. Caprioli, and Raf Van de Plas. Unsupervised machine learning for exploratory data analysis in imaging mass spectrometry. Mass Spectrometry Reviews, 39 0 (3): 0 245–291, October 2019. ISSN 1098-2787. doi:10.1002/mas.21602. URL http://dx.doi.org/10.1002...

    2019 doi

  36. [44]

    An n-glycome tissue atlas of 15 human normal and cancer tissue types determined by maldi-imaging mass spectrometry

    Elizabeth N Wallace, Connor A West, Colin T McDowell, Xiaowei Lu, Evelyn Bruner, Anand S Mehta, Kiyoko F Aoki-Kinoshita, Peggi M Angel, and Richard R Drake. An n-glycome tissue atlas of 15 human normal and cancer tissue types determined by maldi-imaging mass spectrometry. Scie...

    2024

  37. [45]

    Mass spectrometry imaging for spatially resolved multi-omics molecular mapping

    Hua Zhang, Kelly H Lu, Malik Ebbini, Penghsuan Huang, Haiyan Lu, and Lingjun Li. Mass spectrometry imaging for spatially resolved multi-omics molecular mapping. npj Imaging, 2 0 (1): 0 20, 2024

    2024

  38. [46]

    Comparison of dimensionality reduction methods in mass spectra of astrocytoma and glioblastoma tissues

    Evgeny Zhvansky, Anatoly Sorokin, Vsevolod Shurkhay, Denis Zavorotnyuk, Denis Bormotov, Stanislav Pekov, Alexander Potapov, Evgeny Nikolaev, and Igor Popov. Comparison of dimensionality reduction methods in mass spectra of astrocytoma and glioblastoma tissues. Mass Spectrometr...

    2021

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.