pith. machine review for the scientific record. sign in

arxiv: 2604.22364 · v1 · submitted 2026-04-24 · 📊 stat.AP · stat.CO

Recognition: unknown

Tail-Greedy Unbalanced Haar Wavelet Segmentation for Copy Number Alteration Data

Arief Gusnanto, Henry M. Wood, Maharani Ahsani Ummi, Stuart Barber

Pith reviewed 2026-05-08 09:20 UTC · model grok-4.3

classification 📊 stat.AP stat.CO
keywords copy number alterationsunbalanced Haar waveletsegmentationnext-generation sequencingdual-thresholdingcancer genomicsfalse positive reductionshort segment detection
0
0 comments X

The pith

A dual-thresholding strategy in tail-greedy unbalanced Haar wavelets improves detection of short copy number alterations in noisy sequencing data.

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

The paper proposes modifying the tail-greedy unbalanced Haar (TGUHm) method by adding a dual-thresholding strategy for segmenting copy number alteration data from next-generation sequencing. This aims to reduce false positives from spurious spikes while maintaining sensitivity to both short and long segments, which is particularly useful in low-coverage or noisy data. Simulations under Gaussian and heavy-tailed noise show higher true positive rates and lower false positive rates than methods like CBS, HaarSeg, and FDRSeg, with special gains for short segments. Real data application confirms detection of CNAs linked to cancer genes, suggesting practical value for genomic analysis.

Core claim

The central claim is that the TGUHm method with dual-thresholding effectively suppresses spurious spikes in the segmentation process while preserving sensitivity to short and long CNA segments, resulting in superior performance in detecting copy number alterations compared to existing methods.

What carries the argument

The tail-greedy unbalanced Haar wavelet segmentation enhanced with a dual-thresholding strategy, which applies two levels of thresholds to control the detection of segment boundaries and suppress noise-induced artifacts.

If this is right

  • Improved accuracy in identifying short CNA segments under various noise conditions.
  • Lower false positive rates in segmentation of noisy next-generation sequencing data.
  • More reliable identification of cancer-related genes from genomic profiles.
  • Competitive performance across both simulated and real datasets.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The dual-thresholding could generalize to other wavelet-based segmentation tasks beyond CNA detection.
  • Future work might test the method on datasets with different sequencing depths or error profiles.
  • Integration with machine learning could further enhance segment boundary detection.

Load-bearing premise

The noise patterns in new datasets will resemble the Gaussian, heavy-tailed, or single real cancer dataset used in the evaluations, allowing the dual thresholds to balance sensitivity and specificity effectively.

What would settle it

Apply the method to a new cancer dataset with known short CNAs and higher or different noise levels; if short segments are missed or false positives increase significantly beyond the reported rates, the improvement claim would not hold.

Figures

Figures reproduced from arXiv: 2604.22364 by Arief Gusnanto, Henry M. Wood, Maharani Ahsani Ummi, Stuart Barber.

Figure 1
Figure 1. Figure 1: Example of copy number ratio data from one patient, TMA-93. The data was normalised using CNAnorm Gusnanto et al. [2012] and regions with missing values, such as the centromeres, are removed. Each point in the figure denotes the copy number ratio of TMA-93 which corresponds to a specific genomic window (150 kb). 2 Method 2.1 Patients, sequence data and alignments In this study, we considered squamous cell … view at source ↗
Figure 2
Figure 2. Figure 2: The true patterns of copy number alterations, denoted f, in simulated examples. (A) First true function. The irregular pattern of segment length is based on common patterns observed in real data. (B) Second true function, which aims to characterise the proposed method’s performance in a case where the underlying true pattern only contains short altered segments. (C) Third true function. An extreme case whe… view at source ↗
Figure 3
Figure 3. Figure 3: Performance metrics for 1000 replicates of the first test function (see panel A of view at source ↗
Figure 4
Figure 4. Figure 4: Performance metrics for 1000 replicates of the second test function (see panel B of view at source ↗
Figure 5
Figure 5. Figure 5: Performance metrics for 1000 replicates of the third test function (see panel C of view at source ↗
Figure 6
Figure 6. Figure 6: AUC of ROC of the methods applied to the first, second, and third test functions ( view at source ↗
Figure 7
Figure 7. Figure 7: Partial AUC F P < 20 of ROC of the methods applied to the first, second, and third test functions ( view at source ↗
Figure 8
Figure 8. Figure 8: Proportion of times a change-point is estimated against location out of 1000 simulated datasets contaminated with a mixture of two Gaussian distributions 0.95 × N(0, σ2 ) + 0.05 × N(0, 3σ 2 ) for σ 2 = 0.3 2 . The dots denote the proportion of detection at locations where there are actual change-points. The grey solid line is the corresponding test function, repeated here from panel A of view at source ↗
Figure 9
Figure 9. Figure 9: CNA estimate as a result of segmentation of chromosome 8 in patient TMA-93. (A) The copy number ratio data of chromosome 8 in Patient TMA-93. (B) TGUHm segmentation (the detailed location of altered regions are shown on view at source ↗
read the original abstract

Detecting copy number alterations (CNAs) from next-generation sequencing data remains challenging, particularly for short segments under noisy conditions. Existing segmentation methods often suffer from high false positive rates or fail to reliably detect short aberrations, especially in low-coverage data. In this study, we propose a modified tail-greedy unbalanced Haar (TGUHm) method that introduces a dual-thresholding strategy to improve segmentation accuracy. The proposed approach effectively suppresses spurious spikes while preserving sensitivity to both short and long CNA segments. Extensive simulation studies under Gaussian and heavy-tailed noise demonstrate that TGUHm consistently achieves higher true positive rates and lower false positive rates compared to state-of-the-art methods, including CBS, HaarSeg, and FDRSeg. In particular, the proposed method improves detection accuracy for short segments while maintaining competitive overall performance. Application to real cancer genomic data further confirms the practical utility of the method, revealing biologically meaningful CNAs associated with known cancer-related genes. These results suggest that TGUHm provides a robust and effective framework for CNA detection in challenging sequencing settings.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript presents a modified tail-greedy unbalanced Haar (TGUHm) wavelet segmentation method for detecting copy number alterations (CNAs) in next-generation sequencing data. The key innovation is a dual-thresholding strategy designed to reduce false positives from spurious spikes while maintaining sensitivity to short and long CNA segments. Through simulations under Gaussian and heavy-tailed noise, the authors claim superior true positive rates (TPR) and false positive rates (FPR) compared to CBS, HaarSeg, and FDRSeg, with particular improvements for short segments. The method is also applied to a real cancer dataset to identify biologically relevant CNAs.

Significance. If the claimed performance gains hold under broader conditions, TGUHm could provide a useful refinement for CNA detection in noisy or low-coverage NGS settings, where short segments are particularly difficult to resolve. The adaptation of tail-greedy unbalanced Haar wavelets targets a known limitation of standard wavelet and segmentation approaches. The simulation design covering two noise families and the real-data example are positive elements, but the lack of quantitative performance metrics, error bars, and parameter robustness checks limits the immediate impact.

major comments (3)
  1. [Abstract] Abstract: the claim that TGUHm 'consistently achieves higher true positive rates and lower false positive rates' is presented without numerical effect sizes, confidence intervals, or any description of how the dual-threshold values were selected or validated, which is load-bearing for the central performance assertion.
  2. [Simulation studies] Simulation studies: no sensitivity analysis, cross-validation, or justification is given for the specific dual-threshold parameters; this directly affects whether the reported TPR/FPR advantages generalize beyond the exact simulated Gaussian and heavy-tailed models used for evaluation.
  3. [Real data application] Real-data application: only a single cancer dataset is analyzed, with no reported details on data exclusion rules, preprocessing steps, or multiple-testing correction, which weakens the support for the claim of 'practical utility' and 'biologically meaningful CNAs'.
minor comments (2)
  1. [Method] The manuscript would benefit from explicit equations defining the dual-thresholding rule and the precise form of the modified tail-greedy step.
  2. [Figures] Simulation figures should include error bars or variability measures across replicates to allow readers to assess the stability of the reported TPR/FPR differences.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thoughtful comments, which have helped us identify areas for improvement in the presentation of our results. We address each major comment in turn and outline the revisions we intend to make to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim that TGUHm 'consistently achieves higher true positive rates and lower false positive rates' is presented without numerical effect sizes, confidence intervals, or any description of how the dual-threshold values were selected or validated, which is load-bearing for the central performance assertion.

    Authors: We agree that the abstract would benefit from more specific quantitative support for the performance claims. In the revised manuscript, we will update the abstract to include key numerical results from the simulations, such as the TPR and FPR values achieved for short and long segments under both noise models. We will also briefly describe the procedure used to select the dual-threshold values. Full details on these aspects, including any available measures of variability across simulation replicates, will be expanded in the simulation studies section. revision: yes

  2. Referee: [Simulation studies] Simulation studies: no sensitivity analysis, cross-validation, or justification is given for the specific dual-threshold parameters; this directly affects whether the reported TPR/FPR advantages generalize beyond the exact simulated Gaussian and heavy-tailed models used for evaluation.

    Authors: The specific dual-threshold parameters were determined through preliminary simulation experiments aimed at balancing detection sensitivity and specificity, though this process was not fully detailed in the original submission. We concur that a sensitivity analysis would better demonstrate the robustness of our findings. Accordingly, we will add a sensitivity analysis in the revised version, including plots or tables showing how TPR and FPR vary with different threshold settings under the Gaussian and heavy-tailed noise models. This will help confirm that the advantages over competing methods are not overly sensitive to the exact parameter choices. revision: yes

  3. Referee: [Real data application] Real-data application: only a single cancer dataset is analyzed, with no reported details on data exclusion rules, preprocessing steps, or multiple-testing correction, which weakens the support for the claim of 'practical utility' and 'biologically meaningful CNAs'.

    Authors: The real-data example is intended as an illustration of the method's application rather than a comprehensive validation study. We will revise this section to include detailed descriptions of the data preprocessing steps, exclusion criteria applied to the cancer dataset, and our approach to multiple-testing (which relies on the method's thresholding mechanism). While expanding to additional datasets would further support generalizability, we believe the current analysis, with these added details, adequately demonstrates the practical utility in identifying biologically relevant CNAs. revision: partial

Circularity Check

0 steps flagged

No circularity: algorithmic modification evaluated on separate simulations and real data

full rationale

The paper presents TGUHm as an algorithmic extension of the tail-greedy unbalanced Haar wavelet method, introducing a dual-thresholding strategy to suppress spurious spikes while retaining sensitivity to short and long segments. No equations, derivations, or first-principles results are shown that reduce claimed performance metrics to quantities fitted from the same data or defined in terms of the target outputs. Performance claims rest on separate simulation studies (Gaussian and heavy-tailed noise) and application to one real cancer dataset, with comparisons to CBS, HaarSeg, and FDRSeg. Any prior citations to unbalanced Haar work are external to the present modifications and do not form a self-referential chain that forces the reported TPR/FPR improvements. The derivation chain is therefore self-contained and non-circular.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the dual thresholds are implicitly introduced but their selection rule is not stated.

pith-pipeline@v0.9.0 · 5494 in / 1188 out tokens · 43855 ms · 2026-05-08T09:20:59.165573+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

20 extracted references · 1 canonical work pages

  1. [1]

    Correcting for cancer genome size and tumour cell content enables better estimation of copy number alterations from next-generation sequence data

    Arief Gusnanto, Henry Wood, Yudi Pawitan, Pamela Rabbitts, and Stefano Berri. Correcting for cancer genome size and tumour cell content enables better estimation of copy number alterations from next-generation sequence data. Bioinformatics, 28 0 (1): 0 40--47, 2012

    2012

  2. [2]

    Circular binary segmentation for the analysis of array-based dna copy number data

    AB Olshen, ES Venkatraman, Robert Lucito, and Michael Wigler. Circular binary segmentation for the analysis of array-based dna copy number data. Biostatistics, 5 0 (4): 0 557--572, 2004

    2004

  3. [3]

    Efficient change point detection for genomic sequences of continuous measurements

    Vito Muggeo and Giada Adelfio. Efficient change point detection for genomic sequences of continuous measurements. Bioinformatics, 27 0 (2): 0 161--166, 2010

    2010

  4. [4]

    A fast and flexible method for the segmentation of acgh data

    Erez Ben-Yaacov and Yonina C Eldar. A fast and flexible method for the segmentation of acgh data. Bioinformatics, 24 0 (16): 0 i139--i145, 2008

    2008

  5. [5]

    Fdr-control in multiscale change-point segmentation

    Housen Li, Axel Munk, and Hannes Sieling. Fdr-control in multiscale change-point segmentation. Electronic Journal of Statistics, 10 0 (1): 0 918--959, 2016

    2016

  6. [6]

    Copynumber: Efficient algorithms for single- and multi-track copy number segmentation

    Gro Nilsen, Knut Liestøl, Peter Van Loo, Hans Kristian Moen Vollan, Marianne B Eide, Oscar M Rueda, Suet-Feung Chin, Roslin Russell, Lars O Baumbusch, Carlos Caldas, Anne-Lise Børresen-Dale, and Ole Christian Lingjaerde. Copynumber: Efficient algorithms for single- and multi-track copy number segmentation. BMC genomics, 13 0 (1): 0 591--591, 2012

    2012

  7. [7]

    Comparative analysis of algorithms for identifying amplifications and deletions in array cgh data

    Weil R Lai, Mark D Johnson, Raju Kucherlapati, and Peter J Park. Comparative analysis of algorithms for identifying amplifications and deletions in array cgh data. Bioinformatics, 21 0 (19): 0 3763--3770, 2005

    2005

  8. [8]

    A comparison study: Applying segmentation to array CGH data for downstream analyses

    Hanni Willenbrock and Jane Fridlyand. A comparison study: Applying segmentation to array CGH data for downstream analyses. Bioinformatics, 21 0 (22): 0 4084--4091, 2005

    2005

  9. [9]

    Using next-generation sequencing for high resolution multiplex analysis of copy number variation from nanogram quantities of DNA from formalin-fixed paraffin-embedded specimens

    HM Wood, O Belvedere, C Conway, C Daly, R Chalkley, M Bickerdike, C McKinley, P Egan, L Ross, B Hayward, J Morgan, L Davidson, K MacLennan, TK Ong, K Papagiannopoulos, I Cook, DJ Adams, GR Taylor, and P Rabbitts. Using next-generation sequencing for high resolution multiplex analysis of copy number variation from nanogram quantities of DNA from formalin-f...

    2010

  10. [10]

    Estimating optimal window size for analysis of low-coverage next-generation sequence data

    Arief Gusnanto, Charles C Taylor, Ibrahim Nafisah, Henry M Wood, Pamela Rabbitts, and Stefano Berri. Estimating optimal window size for analysis of low-coverage next-generation sequence data. Bioinformatics, 30 0 (13): 0 1823--1829, 2014

    2014

  11. [11]

    Tail-greedy bottom-up data decompositions and fast multiple change-point detection

    P Fryzlewicz. Tail-greedy bottom-up data decompositions and fast multiple change-point detection. Annals of Statistics, 46 0 (6B): 0 3390--3421, 2018

    2018

  12. [12]

    A computational index derived from whole-genome copy number analysis is a novel tool for prognosis in early stage lung squamous cell carcinoma

    Ornella Belvedere, Stefano Berri, Rebecca Chalkley, Caroline Conway, Fabio Barbone, Federica Pisa, Kenneth MacLennan, Catherine Daly, Melissa Alsop, Joanne Morgan, Jessica Menis, Peter Tcherveniakov, Kostas Papagiannopoulos, Pamela Rabbitts, and Henry M Wood. A computational index derived from whole-genome copy number analysis is a novel tool for prognosi...

    2012

  13. [13]

    Fast and accurate short read alignment with burrows–wheeler transform

    Heng Li and Richard Durbin. Fast and accurate short read alignment with burrows–wheeler transform. Bioinformatics, 25 0 (14): 0 1754--1760, 2009

    2009

  14. [14]

    Assembly of microarrays for genome-wide measurement of dna copy number

    Antoine Snijders, Norma Nowak, Richard Segraves, Stephanie Blackwood, Nils Brown, Jeffrey Conroy, Greg Hamilton, Anna Hindle, Bing Huey, Karen Kimura, Sindy Law, Ken Myambo, Joel Palmer, Bauke Ylstra, Jingzhu Yue, Joe Gray, Ajay Jain, Daniel Pinkel, and Donna Albertson. Assembly of microarrays for genome-wide measurement of dna copy number. Nature Genetic...

    2001

  15. [15]

    Gistic2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers

    Craig H Mermel, Steven E Schumacher, Barbara Hill, Matthew L Meyerson, Rameen Beroukhim, and Gad Getz. Gistic2.0 facilitates sensitive and confident localization of the targets of focal somatic copy-number alteration in human cancers. BioMed Central Ltd, 12 0 (4): 0 R41--R41, 2011

    2011

  16. [16]

    A survey of sampling from contaminated distributions

    JW Tukey. A survey of sampling from contaminated distributions. In Contributions to Probability and Statistics: Essays in Honor of Harold Hotelling, pages 448--485, 1960

    1960

  17. [17]

    Performance evaluation of dna copy number segmentation methods

    Morgane Pierre-Jean, Guillem Rigaill, and Pierre Neuvial. Performance evaluation of dna copy number segmentation methods. Briefings in Bioinformatics, 16 0 (4): 0 600--615, 2015

    2015

  18. [18]

    Ensembl 2022

    Fiona Cunningham, James E Allen, Jamie Allen, Jorge Alvarez-Jarreta, M Ridwan Amode, Irina M Armean, Olanrewaju Austine-Orimoloye, Andrey G Azov, If Barnes, Ruth Bennett, Andrew Berry, Jyothish Bhai, Alexandra Bignell, Konstantinos Billis, Sanjay Boddu, Lucy Brooks, Mehrnaz Charkhchi, Carla Cummins, Luca Da Rin Fioretto, Claire Davidson, Kamalkumar Dodiya...

    work page doi:10.1093/nar/gkab1049 2022

  19. [19]

    Hiemstra

    P.S. Hiemstra. Defensins. In Geoffrey J. Laurent and Steven D. Shapiro, editors, Encyclopedia of Respiratory Medicine, pages 7--10. Academic Press, Oxford, 2006. ISBN 978-0-12-370879-3

    2006

  20. [20]

    Amplification of chromosome 8 genes in lung cancer

    Onur Baykara, Burak Bakir, Nur Buyru, Kamil Kaynak, and Nejat Dalay. Amplification of chromosome 8 genes in lung cancer. Journal of cancer, 6 0 (3): 0 270, 2015

    2015