Pith. sign in

REVIEW 3 major objections 6 minor 27 references

CircFormerMoE: An End-to-End Deep Learning Framework for Circular RNA Splice Site Detection and Pairing in Plant Genomes

T0 review · 3 major / 6 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read CircFormerMoE claims circular RNA splice sites and their pairing can be predicted directly from plant genomic DNA, reporting near-99% detection accuracy and 97.35% pairing accuracy across ten species.

desk verdict A sensible two-stage circRNA model whose headline genome-wide discovery claim is unsupported by center-anchored evaluation; worth reviewing with clear revision requirements. read the letter →

arxiv 2507.08542 v1 pith:B7SPARKY submitted 2025-07-11 cs.LG

Tianyou Jiang This is my paper · ORCID
classification cs.LG
keywords circularRNAsplicesitedetectionpairingplantgenomestransformermixtureofexpertsdeeplearninggenomicDNAprediction
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

CircFormerMoE is an attempt to make circular RNA (circRNA) discovery in plants a DNA-sequence prediction problem rather than an RNA-sequencing problem. The paper breaks the task into two stages: detecting individual splice sites along 5001-base genomic windows, then deciding whether pairs of detected sites form a circle by classifying their surrounding sequence contexts. On annotated data from ten plant species it reports near-99% precision and recall for splice-site detection and 97.35% pairing accuracy, with the species-specific mixture-of-experts design adding measurable gains over a shared-backbone baseline. The payoff, if the result holds, is a fast genome-wide screening tool for a class of RNAs whose known abundance is likely underestimated and whose splice signals deviate from canonical GT-AG motifs.

What carries the argument

The load-bearing object is the CircFormerMoE architecture: an input encoder for one-hot DNA (A, C, G, T, N, and M for concatenation), a local encoder of one-dimensional convolutional residual blocks, a global encoder using lightweight multi-head self-attention based on Performer's FAVOR+ random-feature kernel, and species-specific mixture-of-experts heads on a shared backbone. Task 1 uses a per-position 1x1 convolution head to output a splicing score at every base; Task 2 pools the shared representation and passes it through a per-species linear head to classify pairs. The MoE heads with hard routing are what let the model share transferable sequence features across species while specializing each expert; the linear attention is what keeps the 5001-base window and long-range pairing computationally feasible. The two-task decomposition converts the intractable whole-genome problem into local windows plus one classification step.

What would settle it

Run CircFormerMoE on complete chromosomes with a sliding window and compare its predicted novel splice sites and circRNA pairs against independent RNA-seq back-splice-junction calls or RT-PCR on the same plant tissue; a large drop from the reported 99.2% detection precision, or failure of novel candidates to validate, would falsify the DNA-only discovery claim.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central claim is that circRNA formation signals in plants are learnable from raw genomic sequence. CircFormerMoE is presented as an end-to-end framework that first labels every nucleotide as splice-site or not and then classifies candidate site pairs as joined or not, with shared transformer representations and per-species expert heads. The paper reports that CircFormerMoE reaches 99.2% average precision and recall on splice-site detection across ten species, reports scores higher than the SpliceAI baseline by 7.5 to 8.2 percentage points on those metrics, and achieves 97.35% accuracy on splice-site pairing. It also reports that saliency analysis concentrates importance at the junction and reveals polyA/polyT stretches that occur in 43.7% of real sequences near splice sites versus 17.4% expected at random. From this the paper infers that the model can discover previously unannotated circRNAs directly from plant genomes.

Load-bearing premise

That accuracy measured on 5001-base windows centered on known splice sites carries over to sliding across whole chromosomes, where negative positions dominate and the model has not been evaluated.

Editorial extensions

If this is right

  • If the reported accuracy transfers to unseen genomes, circRNA screening in plants can begin from DNA alone and reserve RNA-seq for validation instead of discovery.
  • The species-specific expert design means adding a new plant species requires only a small new head plus fine-tuning, not a full retraining.
  • Because the global encoder scales linearly with sequence length, the architecture is suited to whole-chromosome scans at the reported inference speed of over one million bases per second on a consumer GPU.
  • The detected enrichment of polyA/polyT near splice sites gives a concrete sequence-level feature that can be tested for a role in plant circRNA biogenesis.

Reading between the lines

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

  • The evaluation windows are centered on known splice sites, so the strongest unexamined risk is that genome-wide scanning inflates false positives; a chromosome-scale test with threshold calibration is the natural next experiment.
  • The negative pairs in Task 2 are drawn from different circRNAs; constructing harder negatives from nearby splice sites or from sites on the same transcript would give a stricter measure of pairing specificity.
  • A direct benchmark against RNA-seq back-splicing tools on matched plant samples, measuring overlap and validating novel candidates, would turn the discovery claim into an operational one.
  • If polyA/T motifs are causal rather than correlational, perturbing those stretches near validated splice sites could provide a CRISPR-based test of the model's learned mechanism.
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 / 6 minor

Summary. The manuscript presents CircFormerMoE, a two-stage transformer/MoE model for circular RNA splice site detection (SSD) and pairing (SSP) in 10 plant species. SSD is framed as position-wise classification on 5001 bp windows, and SSP as binary classification on concatenated ±500 bp windows around candidate splice-site pairs. The authors report near-perfect SSD metrics (AP/AR/F1 ≈ 99%) and 97.35% SSP accuracy, and use saliency maps to claim polyA/T importance. The abstract and discussion further claim that the model predicts circRNAs directly from plant genomic DNA and can discover previously unannotated circRNAs, with code and data released.

Significance. If the genome-wide claim were established, this would be a useful contribution: it would enable DNA-only circRNA screening in plants and reduce dependence on RNA-seq. Strengths include the MoE design for cross-species transfer, the multi-species plant dataset, the released code/data/trained models, and the reported inference speed of over one million bases per second. However, the current evaluation does not demonstrate the central discovery claim, because all metrics are computed on windows that are centered on known splice sites rather than on a whole-genome scan.

major comments (3)
  1. [II.A and III.A, Table II, Fig. 4] The SSD evaluation is performed exclusively on 5001 bp windows centered on annotated splice sites, so every positive window contains a true splice site at the center, while negatives are sampled from non-splicing regions. A model can attain high AP/AR by learning to output a peak at the window center for gene-like windows, without learning to locate splice sites at arbitrary positions in a genome-wide sliding-window scan. No whole-chromosome evaluation is reported, so the metrics in Table II and Fig. 4 do not support the abstract's claim of predicting circRNAs directly from genomic DNA or discovering unannotated circRNAs. I request a genome-wide scan evaluation with per-chromosome precision/recall and top-k site recovery, and validation of predicted circRNAs against independent RNA-seq or RT-PCR data.
  2. [II.A and III.B] The SSP negatives are constructed by concatenating splice-site pairs from different circRNAs. These random pairs are not representative of the candidate pairs that arise in a genome-wide scan, and the model may rely on trivial biases such as site-to-site distance or annotation-derived pairing statistics. The paper does not report the distance distribution or genomic context of the negative pairs. Please characterize the negative distribution and evaluate SSP on genome-wide candidate pairs or on held-out pairs from the same experimental pipeline.
  3. [II.D and III.C] The interpretability analysis is self-referential: saliency maps are computed on the training sequences, and the same sequences are then used to report polyA/T enrichment (43.7% versus 17.4% in random sequences). This does not independently validate the model's biological basis. Please compute saliency on held-out or independent sequences and compare polyA/T enrichment against matched genomic background, or explicitly present the saliency results as descriptive only.
minor comments (6)
  1. [Throughout] There are several typos, including 'exprimental', 'seperately', and 'gonomes'; the manuscript should be proofread.
  2. [II.B, Eq. (7)] The gradient notation appears reversed: Eq. (7) writes ϕ_G = ∂x / ∂y but the text and Eq. (8) require ∂y / ∂x; please correct.
  3. [II.B and II.C] Architecture hyperparameters (number of layers, hidden dimension, number of attention heads, random feature count m, learning rate, fine-tuning epochs) are not specified in the paper; the code release mitigates this, but the manuscript should include the key settings for reproducibility.
  4. [III.B, Table III] The perfect 100% scores for Lactuca sativa are suspicious and should be investigated for possible data leakage or label duplication rather than attributed only to 'cleaner annotations'.
  5. [I, reference [11]] Reference [11] concerns intronic branch point constraints and does not appear to support the statement about GT-AG motif usage in human pre-mRNA; please verify the citation.
  6. [Title and Abstract] The term 'end-to-end' may be misleading because the method is presented as two separate subtasks; consider clarifying the pipeline structure.

Circularity Check

1 steps flagged · score 2.0 of 10

No load-bearing circularity: the central SSD/SSP results are held-out and benchmarked against SpliceAI; the only self-referential element is the polyA/T saliency enrichment, which restates the training data rather than independently validating the model.

  1. fitted input called prediction [Section III.C 'Model Interpretability' (method in Section II.D, Eqs. 7-9)]
    "Based on this visualization, we scanned the circRNA sequences in our dataset to identify the presence of obvious polyA or polyT stretches (Continuous ≥ 5 bases) within a 100-base window around the splice site. Compared to theoretical random sequences, where approximately 17.4% of sequences contain such polyA/T regions, our real data showed a markedly higher occurrence. Specifically, 43.7% of sequences in our dataset contained polyA/T stretches within this region."

    The saliency maps are computed on 'all circRNA splicing site sequence across 10 species' (Section III.C, Eq. 9), i.e., the same PlantCircRNA-derived sequences used to train the model. The 43.7% polyA/T frequency is then measured on 'our dataset' — the same training-set sequences — and presented as evidence for the biological relevance of a feature the model learned from that dataset. Both the gradient-based saliency and the enrichment statistic are functions of the same input distribution, so the enrichment does not independently validate the learned feature; it restates a property of the training data. This is a minor self-referential observation, not a load-bearing derivation of the paper's central detection/pairing claims.

full rationale

The paper's central performance claims are not circular. Task 1 (SSD) and Task 2 (SSP) are evaluated on held-out test sequences derived from the same PlantCircRNA database but not used for training, and the headline numbers (AP 0.992, AR 0.992, F1 0.992, top-k 0.919) are compared against an external baseline, SpliceAI, using standard metrics. No fitted parameter from the test set is renamed as a prediction, and no uniqueness theorem or load-bearing self-citation forces the architecture choice. The genome-wide discovery claim ('capable of discovering previously unannotated circRNAs') is indeed unsupported — the paper never runs a whole-chromosome scan and the Discussion only proposes sliding-window inference — but that is an extrapolation/validation gap, not a circular reduction: the model's evaluation windows are centered on known sites, yet the claim is not derived by an equation that makes it equal to the training input. The one genuinely self-referential element is the interpretability analysis: the saliency-weighted logo is computed from a model trained on the same circRNA sequences, and the subsequent polyA/T enrichment (43.7% vs. 17.4% random) is measured on that same dataset. This makes the 'biologically relevant features' conclusion a restatement of training-data statistics rather than an independent external confirmation. This does not undermine the held-out SSD/SSP metrics, which retain independent content. Overall circularity is therefore minor: score 2.

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

The central claim relies on the PlantCircRNA database as ground truth, on the negative sampling scheme, and on the assumption that snippet-level evaluation transfers to whole-genome scanning. The class weights and window lengths are hand-chosen without sensitivity analysis. No new physical entities are introduced.

free parameters (4)
  • SSD input window length = 5001 bp
    Chosen as 2500 bases on each side of a splice site; no sensitivity analysis or ablation is reported, and it determines the context available for detection.
  • SSP context window length = 1001 bp (plus 500 bp upstream and downstream)
    Described as empirically selected as a trade-off between accuracy and speed; the exact choice affects Task 2 classification results.
  • Positive class weight for SSD = 285 for the first 10 epochs
    Set based on the ratio of negative to positive positions in the training data; later removed, but it strongly shapes early training and the reported learning curves.
  • Positive class weight for SSP = 10 for the first 5 epochs
    Set based on the 10:1 negative to positive sampling ratio; a hand-chosen training schedule that could affect final metrics.
assumptions (4)
  • domain assumption PlantCircRNA database annotations are accurate ground-truth labels for circular RNA splice sites.
    All training and test labels come from this database (Section II.A); no independent experimental verification of the annotations is provided, and errors in the database would directly bias the reported metrics.
  • domain assumption Negative samples, non-splicing genomic windows in Task 1 and random site pairs in Task 2, are representative of true genome-wide negatives.
    Random pairs of splice sites from different circRNAs may not model the real distribution of non-circular pairs, and non-splicing windows are not clearly defined (Section II.A). If these negatives are too easy, the model may not generalize.
  • standard math The Performer random feature map (FAVOR+) provides a sufficiently accurate approximation of softmax attention for this task.
    The paper relies on the known linear attention approximation (Section II.B.3) without ablating its effect on results, but this is a standard technique with established bounds.
  • domain assumption SpliceAI can be fairly fine-tuned for this binary splice-site task, despite being designed for donor/acceptor classification in human pre-mRNA.
    No details are given for how SpliceAI was adapted or fine-tuned (Section III.A), yet it is the main external baseline. If the adaptation is suboptimal, the comparison is unfair.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CircFormerMoE: An End-to-End Deep Learning Framework for Circular RNA Splice Site Detection and Pairing in Plant Genomes." pith.science (2026). https://pith.science/paper/B7SPARKY

@misc{pith2026250708542,
  author       = {Pith},
  title        = {Pith review of: CircFormerMoE: An End-to-End Deep Learning Framework for Circular RNA Splice Site Detection and Pairing in Plant Genomes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B7SPARKY}},
  note         = {Machine review of arXiv:2507.08542}
}
read the original abstract

Circular RNAs (circRNAs) are important components of the non-coding RNA regulatory network. Previous circRNA identification primarily relies on high-throughput RNA sequencing (RNA-seq) data combined with alignment-based algorithms that detect back-splicing signals. However, these methods face several limitations: they can't predict circRNAs directly from genomic DNA sequences and relies heavily on RNA experimental data; they involve high computational costs due to complex alignment and filtering steps; and they are inefficient for large-scale or genome-wide circRNA prediction. The challenge is even greater in plants, where plant circRNA splice sites often lack the canonical GT-AG motif seen in human mRNA splicing, and no efficient deep learning model with strong generalization capability currently exists. Furthermore, the number of currently identified plant circRNAs is likely far lower than their true abundance. In this paper, we propose a deep learning framework named CircFormerMoE based on transformers and mixture-of experts for predicting circRNAs directly from plant genomic DNA. Our framework consists of two subtasks known as splicing site detection (SSD) and splicing site pairing (SSP). The model's effectiveness has been validated on gene data of 10 plant species. Trained on known circRNA instances, it is also capable of discovering previously unannotated circRNAs. In addition, we performed interpretability analyses on the trained model to investigate the sequence patterns contributing to its predictions. Our framework provides a fast and accurate computational method and tool for large-scale circRNA discovery in plants, laying a foundation for future research in plant functional genomics and non-coding RNA annotation.

Figures

Figures reproduced from arXiv: 2507.08542 by the authors.

Figure 1
Figure 1. The distance distribution of paired circRNA splicing sites, also known as circRNA length among these 10 species is as follows: short distance (<1000 bases): 106626 samples for 82%; Medium distance (1000-5000 base pairs): 8788 samples for 6.8%; Long distance (5000-10000 base pairs): 4001 samples for 3.1%; Ultra long distance (>10000 bases): 10563 samples for 8.1%. Detailed information of datasets used for training an… view at source ↗
Figure 2
Figure 2. CircFormerMoE Architecture. It consists of an input encoder, a local encoder, a global encoder and specific Mixture-of-Experts (MoE) heads for each species. 1) Model Architecture Overview The overall architecture of CircFormerMoE is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Performance comparison between CircFormerMoE and CircFormer trained across all species. The left panel shows the mean average recall across the 10 species in the evaluation set, while the right panel presents the mean average precision. The values are averaged from diverse results of 10 species. TABLE II SSD EVALUATION METRICS AVERAGED ACROSS EVALUATION SETS OF 10 SPECIES. Model AP (%) AR (%) F1 (%) Top-k acc (%) Ci… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Detailed evaluation results on Task 1 across 10 species. Each subplot shows a comparison among three models including CircFormerMoE, CircFormer and SpliceAI, on one of the following metrics: precision (top), recall (middle), and top-k accuracy (bottom). B. Splicing Sit…
Figure 5
Figure 5. Figure 5: Heatmap of evaluation metrics on Task 2 across 10 species. Each row corresponds to a species and each column to a metric on the test set. The color intensity reflects the performance score, with higher values indicating better performance. C. Model Interpretability To …
Figure 6
Figure 6. Figure 6: Saliency map of position importance of the model prediction surrounding circularRNA splicing sites, where the middle position corresponds to the splice sites [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Saliency-Weighted Sequence Logo represents the relative importance of nucleotides within a ±50 bp window centered on the splice site, averaged across all species. Each position shows the normalized contribution of adenine (A), cytosine (C), guanine (G), and thymine (T)…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 25 canonical work pages

  1. [1]

    Cell- type specific features of circular RNA expression,

    J. Salzman, R. E. Chen, M. N. Olsen, P. L. Wang, and P. O. rown, “Cell- type specific features of circular RNA expression,” PLoS Genet., vol. 9, p. e1003777, 2013

    work page 2013

  2. [2]

    circRNA biogenesis competes with pre -mRNA splicing,

    R. Ashwal -Fluss et al., “circRNA biogenesis competes with pre -mRNA splicing,” Mol. Cell, vol. 56, pp. 55–66, 2014

    work page 2014

  3. [3]

    Recent research progress on circular RNAs: iogenesis, properties, functions, and therapeutic potential,

    A. R. Sharma, M. Bhattacharya, S. Bhakta, A. Saha, S. S. Lee, and C. Chakraborty, “Recent research progress on circular RNAs: iogenesis, properties, functions, and therapeutic potential,” Mol. Ther. Nucleic Acids, vol. 25, pp. 355–371, 2021

    work page 2021

  4. [4]

    Circular RNAs: Promising molecular biomarkers of human aging -related diseases via functioning as an miRNA sponge,

    S. Ren, P. Lin, J. Wang, H. Yu, T. Lv, L. Sun, and G. Du, “Circular RNAs: Promising molecular biomarkers of human aging -related diseases via functioning as an miRNA sponge,” Mol. Ther. Methods Clin. Dev., vol. 18, pp. 215–229, 2020

    work page 2020

  5. [5]

    CircRNAs and cancer: iomarkers and master regulators,

    E. Arnaiz, C. Sole, L. Manterola, L. Iparraguirre, D. Otaegui, and C. H. Lawrie, “CircRNAs and cancer: iomarkers and master regulators,” in Semin. Cancer Biol., vol. 58, pp. 90–99, Oct. 2019

    work page 2019

  6. [6]

    The emerging roles and functions of circular RNAs and their generation,

    C. Y. Yu and H. C. Kuo, “The emerging roles and functions of circular RNAs and their generation,” J. iomed. Sci., vol. 26, no. 1, p. 29, 2019

    work page 2019

  7. [7]

    Circular RNA - protein interactions: Functions, mechanisms, and identification,

    A. Huang, H. Zheng, Z. Wu, M. Chen, and Y. Huang, “Circular RNA - protein interactions: Functions, mechanisms, and identification,” Theranostics, vol. 10, no. 8, pp. 3503–3517, 2020

    work page 2020

  8. [8]

    Diverse alternative back -splicing and alternative splicing landscape of circular RNAs,

    X. O. Zhang et al., “Diverse alternative back -splicing and alternative splicing landscape of circular RNAs,” Genome Res., vol. 26, no. 9, pp. 1277–1287, 2016

    work page 2016

Show all 27 references
  1. [9]

    Circular RNAs are a large class of animal RNAs with regulatory potency,

    S. Memczak et al., “Circular RNAs are a large class of animal RNAs with regulatory potency,” Nature, vol. 495, no. 7441, pp. 333–338, 2013

    2013

  2. [10]

    CircRNAs in plants,

    X. Lai, J. Bazin, S. Webb, M. Crespi, C. Zubieta, and S. J. Conn, “CircRNAs in plants,” in CircRNAs: iogenesis and Functions, 2018, pp. 329–343

    2018

  3. [11]

    The intronic branch point se uence is under strong evolutionary constraint,

    L. F. Lareau et al., “The intronic branch point se uence is under strong evolutionary constraint,” Nat. Commun., vol. 12, no. 1, p. 1, 2021

    2021

  4. [12]

    Present scenario of circular RNAs (circRNAs) in plants,

    W. Zhao, S. Chu, and Y. Jiao, “Present scenario of circular RNAs (circRNAs) in plants,” Front. Plant Sci., vol. 10, p. 379, 2019

    2019

  5. [13]

    Characterization and function of circular RNAs in plants,

    P. Zhang, S. Li, and M. Chen, “Characterization and function of circular RNAs in plants,” Front. Mol. iosci., vol. 7, p. 91, 2020

    2020

  6. [14]

    Circular RNA discovery with emerging se uencing and deep learning technologies,

    J. Zhang and F. Zhao, “Circular RNA discovery with emerging se uencing and deep learning technologies,” Nat. Genet., pp. 1–14, 2025

    2025

  7. [15]

    Predicting splicing from primary se uence with deep learning,

    K. Jaganathan et al., “Predicting splicing from primary se uence with deep learning,” Cell, vol. 176, no. 3, pp. 535–548, 2019

    2019

  8. [16]

    The implementation of deep learning algorithm with Gaussian blur data preprocessing in circular RNA classification and detection,

    E. Leovonzko, C. F. Cahyaningrum, and R. Ulwani, “The implementation of deep learning algorithm with Gaussian blur data preprocessing in circular RNA classification and detection,” Undergrad. Res. Nat. Clin. Sci. Technol. J., vol. 8, pp. 1–10, 2024

    2024

  9. [17]

    AScirRNA: A novel computational approach to discover abiotic stress -responsive circular RNAs in plant genome,

    U. K. Pradhan et al., “AScirRNA: A novel computational approach to discover abiotic stress -responsive circular RNAs in plant genome,” Comput. Biol. Chem., vol. 113, p. 108205, 2024

    2024

  10. [18]

    DeepECD: A model for predicting plant extrachromosomal circular DNA from se uences,

    J. Hu, Z. Tang, Y. Wang, J. Yan, and X. Sun, “DeepECD: A model for predicting plant extrachromosomal circular DNA from se uences,” J. Biotech Res., vol. 16, 2024

    2024

  11. [19]

    PlantCircRNA: A comprehensive database for plant circular RNAs,

    S. He et al., “PlantCircRNA: A comprehensive database for plant circular RNAs,” Nucleic Acids Res., vol. 53, no. D1, pp. D1595–D1605, 2025

    2025

  12. [20]

    Ensembl 2021,

    K. L. Howe et al., “Ensembl 2021,” Nucleic Acids Res., vol. 49, no. D1, pp. D884–D891, 2021

    2021

  13. [21]

    Transformers are RNNs: Fast autoregressive transformers with linear attention,

    A. Katharopoulos, A. Vyas, N. Pappas, and F. Fleuret, “Transformers are RNNs: Fast autoregressive transformers with linear attention,” in Proc. nt. Conf. Mach. Learn., Nov. 2020, pp. 5156–5165

    2020

  14. [22]

    Attention is all you need,

    A. Vaswani et al., “Attention is all you need,” in Adv. Neural nf. Process. Syst., vol. 30, 2017

    2017

  15. [23]

    Rethinking attention with performers,

    K. Choromanski et al., “Rethinking attention with performers,” arXiv preprint arXiv:2009.14794, 2020

    2009 arXiv

  16. [24]

    Deep inside convolutional networks: Visualising image classification models and saliency maps,

    K. Simonyan, A. Vedaldi, and A. Zisserman, “Deep inside convolutional networks: Visualising image classification models and saliency maps,” arXiv preprint arXiv:1312.6034, 2013

    2013 arXiv

  17. [25]

    Se uence logos: a new way to display consensus se uences,

    T. D. Schneider and R. M. Stephens, “Se uence logos: a new way to display consensus se uences,” Nucleic Acids Res., vol. 18, no. 20, pp. 6097–6100, 1990

    1990

  18. [26]

    Logomaker: beautiful se uence logos in Python,

    A. Tareen and J. . Kinney, “Logomaker: beautiful se uence logos in Python,” ioinformatics, vol. 36, no. 7, pp. 2272–2274, 2020

    2020

  19. [27]

    H., Mao, A., Salzberg, S

    Chao, K. H., Mao, A., Salzberg, S. L., & Pertea, M. (2024). Splam: a deep-learning-based splice site predictor that improves spliced alignments. Genome Biology, 25(1), 243

    2024

Pith tools

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