Pith. sign in

REVIEW 5 major objections 5 minor 43 references

MERGE: Multi-faceted Hierarchical Graph-based GNN for Gene Expression Prediction from Whole Slide Histopathology Images

T0 review · 5 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read MERGE predicts per-spot gene expression by clustering tissue spots by both location and visual morphology; on ST-Net its Pearson correlation reaches 0.6795, up from 0.5789 with one-hop edges and 0.2320 for TRIPLEX.

desk verdict The hierarchical graph is a genuine step forward, but the SPCS-smoothed evaluation target likely inflates the reported gains, and the baseline reproductions need verification. read the letter →

arxiv 2412.02601 v2 pith:CZPTYTH2 submitted 2024-12-03 cs.CV

Aniruddha Ganguly , Debolina Chatterjee , Wentao Huang , Jie Zhang , Alisa Yurovsky , Travis Steele Johnson , Chao Chen This is my paper · ORCID
classification cs.CV
keywords spatialtranscriptomicsgeneexpressionpredictionwholeslideimagesgraphneuralnetworkshierarchicalconstructionhistopathologySPCSsmoothinggene-awaredata
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 sets out to show that gene-expression profiles at every spot of a tissue slide can be predicted from the histology image alone, provided the predictor is allowed to pool information across spots that are either close in tissue space or similar in appearance. It presents MERGE, which clusters image patches twice—once by physical coordinates and once by image-feature embeddings—then wires each spot to its two cluster centroids and the centroids to one another, so a graph neural network can pass messages between distant but morphologically related regions within a few hops. The reported results on three cancer spatial-transcriptomics datasets (breast and skin) place MERGE ahead of all compared baselines on mean squared error, mean absolute error, and Pearson correlation, with ST-Net Pearson correlation at 0.6795 versus 0.5789 for the same GNN with one-hop edges and 0.2320 for the reproduced TRIPLEX baseline. The paper further argues that the preprocessing choice matters: a gene-aware, space-aware smoothing (SPCS) produces ground-truth expression maps that better match tissue morphology than the eight-neighbor average used by prior work, and the reported superiority is measured against those SPCS-smoothed targets.

What carries the argument

The load-bearing object is the multi-faceted hierarchical graph, built from two parallel clusterings of the $n$ spots in a slide: one in physical space using spot coordinates, and one in feature space using 256-dimensional patch embeddings from a fine-tuned ResNet18 encoder. For each of the $c$ clusters in either clustering, the spot nearest the feature-space centroid serves as the cluster centroid; every spot is joined to both of its centroids (internal edges), the $2c$ centroids form a complete graph among themselves (shortcut edges), and the eight-neighbor spatial edges are kept as well. A four-layer Graph Attention Network with eight attention heads performs message passing on this graph and produces a 250-gene expression vector at each node; the guaranteed three-hop connectivity between any two nodes is what allows short-range spatial context and long-range morphological similarity to be fused in a shallow network.

What would settle it

Re-run the full comparison (MERGE, the one-hop GNN, and TRIPLEX) with raw unsmoothed transcript counts as ground truth instead of SPCS-smoothed values; if the PCC ordering and the size of the gaps stay the same, the smoothing is not the driver, while a sharp shrinkage would tie the claimed superiority to the smoothing step rather than to true expression.

Watch

Extended reading notes

Core claim

MERGE's central claim is that joint prediction of all spots' gene expressions from histology images is improved by a graph whose edges encode two distinct notions of biological relevance. Spots are clustered once by tissue coordinates and once by image-feature (morphology) embeddings; within each cluster every spot is connected to a chosen centroid spot, all centroids are connected among themselves by shortcut edges, and the usual eight-neighbor spatial edges are retained. The resulting graph stays sparse—$O(n)$ internal edges plus $O(c^2)$ shortcut edges—yet guarantees that any two spots can exchange messages in at most three hops, letting the GNN propagate morphology-informed information between distant locations. On this graph a four-layer Graph Attention Network outputs a 250-gene expression vector per node, and the authors report that it outperforms the per-spot ResNet+FCN, the one-hop GNN of Hist2ST, and the multi-resolution TRIPLEX model on MSE, MAE, and Pearson correlation across ST-Net, Her2ST, and SCC. The paper also credits part of this success to adopting SPCS smoothing, a two-factor, transcriptome-aware preprocessing, as the source of the ground-truth expression targets.

Load-bearing premise

The load-bearing premise is that SPCS-smoothed expression—each spot's value replaced by a weighted blend of nearby and expression-similar spots—is the correct ground truth for training and evaluation, so the reported gains may partly measure how well the graph reproduces that smoothing field rather than how well it captures the underlying biology.

Editorial extensions

If this is right

  • On all three benchmarks (ST-Net, Her2ST, SCC), MERGE with the hierarchical graph reports the highest Pearson correlation among the methods compared, lifting ST-Net PCC from 0.5789 with one-hop edges to 0.6795.
  • Because any two spots can exchange information within three hops, a four-layer GAT is enough; very deep networks are not needed to propagate context across a slide.
  • At inference the pipeline is inductive: given any whole-slide image, patches are extracted, the same two clusterings and edge sets are constructed, and all spots receive joint gene-expression predictions without retraining.
  • The smoothing step is not neutral: SPCS targets improve MERGE's PCC from 0.6645 to 0.679 relative to eight-neighbor averaging, while the same switch lowers TRIPLEX's PCC from 0.3014 to 0.2320.

Reading between the lines

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

  • A testable extension the paper does not run: retraining MERGE and the baselines on raw, unsmoothed transcript counts, or on a single-cell-resolved spatial assay, would show whether the graph's advantage reflects genuine biological signal or the recovery of the SPCS smoothing field.
  • The same two-clustering, centroid, and shortcut-edge recipe transfers to other whole-slide prediction problems—survival, grade, or mutation-status classification—where spatially distant but morphologically similar regions are plausibly informative; that transfer is an open experiment.
  • Because the feature-space clustering consumes any patch embeddings, MERGE's graph construction can sit on top of newer foundation-model encoders, and whether the reported gains persist with stronger features is untested.
  • A head-to-head comparison the paper does not perform is a graph transformer given the identical hierarchical edges; if its accuracy matches the GAT, the hierarchy rather than attention would carry the result.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes MERGE, a graph neural network for predicting per-spot gene expression from whole-slide histopathology images. The method builds a sparse multi-faceted hierarchical graph from two clusterings (spatial coordinates and image-feature embeddings), connects each cluster to a centroid spot, adds complete shortcut edges among centroids, and retains the 8-neighbor spatial edges; a four-layer GAT then predicts 250 gene expression values per spot. The authors also advocate replacing the commonly used 8-neighbor spatial smoothing with SPCS, a two-factor smoother combining spatial distance and gene-expression pattern similarity. Experiments on ST-Net, Her2ST, and SCC with eight-fold cross-validation report PCC, MSE, and MAE against six baselines, with internal ablations supporting the contribution of each graph component.

Significance. The multi-faceted graph construction is a sensible and potentially useful inductive bias for spatial transcriptomics: it explicitly creates short- and long-range edges between spots that are spatially close or morphologically similar, and the internal ablation in Table 2 is coherent (PCC rises from 0.5789 with 1-hop edges to 0.6795 with the full hierarchical graph). If the evaluation concerns below are resolved, the method would be a competitive contribution and the SPCS analysis could be a useful practical note. The paper does not provide code, and the current evaluation makes the claimed superiority over prior work not yet convincing because of the smoothed evaluation target, the modified TRIPLEX baseline, and the absence of uncertainty quantification.

major comments (5)
  1. [Sec. 5 (Pre-processing), Sec. 4, Sec. 3.3] The evaluation target is the SPCS-smoothed gene expression matrix, and SPCS constructs each spot's smoothed value as a weighted combination over essentially all other spots using two factors: spatial distance and gene-expression pattern similarity (PCA plus Pearson correlation). MERGE's graph and features are built from the same two factors: spatial clustering, feature-space clustering, and a patch encoder fine-tuned on the same SPCS-smoothed targets. The reported gains over 1-hop graphs (Table 2) and over baselines may therefore capture the model's ability to reproduce the smoothing field rather than to predict true gene expression. This is a correctness-risk concern, not a statement that the method is invalid. Concrete tests: report results with raw (or 8n-smoothed) expression as the training and evaluation target, and report results with a patch encoder that has not been fine-tuned on gene-expression targets. The 8n row of Table 4 partially mitigates the concern (MERGE PCC 0.6645 vs TRIPLEX 0.3014), but the internal graph-construction gain is not shown on non-SPCS targets, and the biological-interpretation claims in Sec. 5.2 are all made against the smoothed field.
  2. [Sec. 5 (Baselines), Table 1, Sec. 3.3] The TRIPLEX baseline appears to be a modified variant. TRIPLEX is described in the related work and in Sec. 3.3 as a multi-resolution method that extracts features from the target patch, its spatial neighborhood, and the whole slide, and the paper says it uses 'the same ResNet18 architecture as our own patch encoder' in TRIPLEX for fair comparison. Replacing TRIPLEX's multi-resolution encoders with a single-scale ResNet18 removes the mechanism that defines the method. The reported TRIPLEX numbers (e.g., PCC 0.2320 on ST-Net vs 0.579 for 1-hop MERGE in Table 1) should therefore be labeled as a re-implementation with a different backbone, or the original multi-resolution TRIPLEX should be run on the same smoothed data. Without this, the SOTA comparison is not established.
  3. [Supplementary Sec. 7.1, Tables 1-4] No uncertainty quantification is reported. The text mentions eight-fold cross-validation and five reruns, but the supplementary material states that a constant seed is set and 'the best model is picked for each experiment'. Choosing the best of five runs removes any measure of run-to-run variability and can inflate reported values through selection. The key comparisons in Table 2 (PCC 0.6719 for 'w/o feature' vs 0.6795 for full MERGE) and across datasets in Table 1 are small in some cases, and without standard deviations or paired significance tests the superiority claims are not statistically grounded. Please report mean and standard deviation over folds and reruns under a fixed protocol, or clearly motivate the best-of-five selection.
  4. [Abstract, Table 1 (SCC row)] The abstract's claim that the method 'outperforms state-of-the-art techniques across multiple metrics' is contradicted by the SCC results in Table 1: TRIPLEX has lower MSE (0.4891 vs 0.5353) and lower MAE (0.5356 vs 0.5838), while MERGE has only a slightly higher PCC (0.5512 vs 0.5416). The claim should be qualified to specific datasets and metrics, or the SCC discussion should directly address why MSE and MAE are worse for MERGE.
  5. [Table 4, Sec. 4] The paper presents SPCS smoothing as a separate contribution and advocates adopting it, but Table 4 shows that SPCS is not uniformly beneficial: for TRIPLEX, switching from 8n to SPCS decreases PCC from 0.3014 to 0.2320 and increases MSE from 0.0760 to 0.1472. The recommendation to use SPCS therefore depends on the model, and the paper should discuss the conditions under which gene-aware smoothing helps or hurts rather than presenting it as a general improvement.
minor comments (5)
  1. [Sec. 3.1] In the sentence introducing shortcut edges, 'facility long-range communication' should be 'facilitate long-range communication'.
  2. [Supplementary Sec. 8.1, Figs. 14-19] The supplementary text refers to the 'FNAS' gene in several places; the gene name is FASN, and this typo should be corrected throughout.
  3. [Fig. 5, Sec. 4] The comparison in Fig. 5 is stated to be 'directly taken from the original SPCS paper', but the caption does not make this provenance clear to the reader; if this is a reproduction rather than an original experiment, the caption should say so explicitly.
  4. [Sec. 5.3, Table 3] The cluster-size ablation is performed only on ST-Net, and the text does not state whether the chosen cluster size (100) is then used for Her2ST and SCC without further tuning; this should be clarified so the reader can judge whether the reported Her2ST and SCC results involve any dataset-specific selection.
  5. [Sec. 5 (Baselines)] The text does not describe the hyperparameter search or training budget for the baselines; reporting these details would help the reader assess whether the comparisons are balanced.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the graph uses spatial coordinates and supervised image embeddings, evaluated by cross-validation; SPCS smoothing is a shared protocol choice, not a fitted input.

full rationale

MERGE's derivation chain is self-contained. The multi-faceted hierarchical graph is constructed from (i) tissue position coordinates and (ii) patch embeddings produced by a ResNet18 encoder that is fine-tuned on gene-expression targets; at inference the graph uses only image-derived features and coordinates, never the test expression values (Sec. 3.1, 3.3, 5). The claimed gain of the hierarchical graph over 1-hop edges and prior methods (Tab. 1: PCC 0.5789 to 0.6795 on ST-Net; Tab. 2 ablations) is an empirical comparison under the same encoder and protocol, not a quantity that equals a fitted parameter by construction. The use of SPCS-smoothed expression as the shared evaluation target is a preprocessing choice; although SPCS's two weighting factors (spatial distance and gene-expression-pattern similarity) are conceptually mirrored by MERGE's spatial and feature-space clustering facets, the paper does not define the graph in terms of the smoothed expression values, and every baseline is trained and tested on the same target. The citation of SPCS (Ref. [26]) is code-reproduced in the paper with stated parameters (Sec. 7.2, Tab. 5) and is not load-bearing for the central GNN claim. The SPCS-vs-8n comparison (Tab. 4) changes the target while changing the smoothing, which is a validity concern rather than a self-definitional or fitted-input reduction. No load-bearing self-citation, imported uniqueness claim, ansatz-by-citation, or renamed known result was found.

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

The target-side degrees of freedom dominate: SPCS smoothing converts raw counts into a globally smoothed field used as both training target and evaluation ground truth, and the cluster size is tuned on the reported dataset. Model-side parameters are standard GAT and encoder hyperparameters. The assumptions are biologically plausible morphology-expression priors plus the evaluation premise that smoothed expression is the right ground truth. No invented entities are introduced; the hierarchical graph is a model component with direct empirical support.

free parameters (5)
  • Cluster size (spatial and feature clusterings) = 100
    Ablated on ST-Net (Tab. 3); used for all reported results. Performance peaks at 100 and degrades beyond.
  • SPCS smoothing parameters = tau_s=2, tau_p=16, alpha=0.6, beta=0.4
    Adopted from the SPCS paper (Tab. 5); they define the smoothed target used for training and evaluation.
  • GAT depth, attention heads, edge dropout = 4 layers, 8 heads, p=0.2
    Architecture choices in Sec. 3.2, stated without ablation.
  • Patch embedding dimension = 256
    Penultimate layer size of the fine-tuned ResNet18 (Sec. 3.3); feeds the feature-space clustering.
  • Gene target subset = 250 genes per dataset
    Inherited from the TRIPLEX preprocessing; defines the prediction target.
assumptions (5)
  • domain assumption Spatially adjacent spots have similar morphology and correlated gene expression.
    Justifies spatial clustering and 1-hop edges in Sec. 3.1.
  • domain assumption Spots with similar image features share morphology and therefore correlated gene expression.
    Premise for feature-space clustering (Sec. 3.1); the features are fine-tuned on the target, so this is partly learned rather than purely morphological.
  • domain assumption SPCS-smoothed expression values are a valid training target and evaluation ground truth.
    Invoked in Secs. 4 and 5; Table 4 shows the smoothing choice changes model rankings, so this premise is load-bearing.
  • ad hoc to paper A single centroid spot can represent its cluster for message passing.
    Sec. 3.1 connects every spot to the spot nearest the feature centroid; ablated but not derived.
  • domain assumption The clustering algorithm yields morphologically meaningful groups at the chosen cluster size.
    Sec. 3.1; the algorithm is unspecified and cluster size is tuned post hoc on ST-Net (Tab. 3).

how reviews work

0 comments
Cite this review

Pith. "Pith review of MERGE: Multi-faceted Hierarchical Graph-based GNN for Gene Expression Prediction from Whole Slide Histopathology Images." pith.science (2026). https://pith.science/paper/CZPTYTH2

@misc{pith2026241202601,
  author       = {Pith},
  title        = {Pith review of: MERGE: Multi-faceted Hierarchical Graph-based GNN for Gene Expression Prediction from Whole Slide Histopathology Images},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CZPTYTH2}},
  note         = {Machine review of arXiv:2412.02601}
}
read the original abstract

Recent advances in Spatial Transcriptomics (ST) pair histology images with spatially resolved gene expression profiles, enabling predictions of gene expression across different tissue locations based on image patches. This opens up new possibilities for enhancing whole slide image (WSI) prediction tasks with localized gene expression. However, existing methods fail to fully leverage the interactions between different tissue locations, which are crucial for accurate joint prediction. To address this, we introduce MERGE (Multi-faceted hiErarchical gRaph for Gene Expressions), which combines a multi-faceted hierarchical graph construction strategy with graph neural networks (GNN) to improve gene expression predictions from WSIs. By clustering tissue image patches based on both spatial and morphological features, and incorporating intra- and inter-cluster edges, our approach fosters interactions between distant tissue locations during GNN learning. As an additional contribution, we evaluate different data smoothing techniques that are necessary to mitigate artifacts in ST data, often caused by technical imperfections. We advocate for adopting gene-aware smoothing methods that are more biologically justified. Experimental results on gene expression prediction show that our GNN method outperforms state-of-the-art techniques across multiple metrics.

Figures

Figures reproduced from arXiv: 2412.02601 by the authors.

Figure 1
Figure 1. Combining image feature-space clustering with spatial [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. The schematic of MERGE shows the overall workflow of our method. (a) Outlines the architecture of our method. The [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The two columns show the original and smoothed ex [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (15 more)
Figure 5
Figure 5. Figure 5: SPCS improves gene expression correspondence with [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: The figure shows the predictions for two cancer-relevant [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: The figure shows clustering outputs for two samples in [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: This presents a bar chart of the sample-by-sample drop [PITH_FULL_IMAGE:figures/full_fig_p011_8.png]
Figure 9
Figure 9. Figure 9: List of the 250 genes used in all experiments on the ST-Net dataset. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png]
Figure 10
Figure 10. Figure 10: List of the 250 genes used in all experiments on the Her2ST dataset. [PITH_FULL_IMAGE:figures/full_fig_p012_10.png]
Figure 11
Figure 11. Figure 11: List of the 250 genes used in all experiments on the SCC dataset. [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]
Figure 12
Figure 12. Figure 12: The figure shows the WSI, extracted patches, outputs of feature space clustering, and outputs of gene space clustering using [PITH_FULL_IMAGE:figures/full_fig_p014_12.png]
Figure 13
Figure 13. Figure 13: When we plot the expression values for two cancer [PITH_FULL_IMAGE:figures/full_fig_p015_13.png]
Figure 14
Figure 14. Figure 14: The PCC between the ground truth and predicted gene expressions for the tumor marker FNAS gene and breast cancer biomarker [PITH_FULL_IMAGE:figures/full_fig_p016_14.png]
Figure 15
Figure 15. Figure 15: This is a histogram of PCC for the two methods for the ST-Net dataset. The upper panel shows the histogram for the FASN [PITH_FULL_IMAGE:figures/full_fig_p017_15.png]
Figure 16
Figure 16. Figure 16: Figure shows PCC between ground truth expressions and predictions for the gene FASN in a few samples. Each row represents [PITH_FULL_IMAGE:figures/full_fig_p018_16.png]
Figure 17
Figure 17. Figure 17: Extension of Fig [PITH_FULL_IMAGE:figures/full_fig_p019_17.png]
Figure 18
Figure 18. Figure 18: Figure shows PCC between ground truth expressions and predictions for the gene GNAS in a few samples. Each row represents [PITH_FULL_IMAGE:figures/full_fig_p020_18.png]
Figure 19
Figure 19. Figure 19: Extension of Fig [PITH_FULL_IMAGE:figures/full_fig_p021_19.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

43 extracted references · 36 canonical work pages

  1. [1]

    Principal component analysis

    Herv ´e Abdi and Lynne J Williams. Principal component analysis. Wiley interdisciplinary reviews: computational statistics, 2(4):433–459, 2010. 6

    work page 2010

  2. [2]

    Goodwin, Anubhav Sinha, et al

    Shahar Alon, Daniel R. Goodwin, Anubhav Sinha, et al. Ex- pansion sequencing: Spatially precise in situ transcriptomics in intact biological systems. Science, 371(6528):eaax2656,

    work page

  3. [3]

    Spatial deconvolution of her2-positive breast tumors reveals novel intercellular relationships

    Alma Andersson, Ludvig Larsson, Linnea Stenbeck, et al. Spatial deconvolution of her2-positive breast tumors reveals novel intercellular relationships. bioRxiv, pages 2020–07,

    work page 2020

  4. [4]

    Layer normalization

    Jimmy Lei Ba. Layer normalization. arXiv preprint arXiv:1607.06450, 2016. 5

    arXiv 2016

  5. [5]

    Boettiger, Jeffrey R

    Kok Hao Chen, Alistair N. Boettiger, Jeffrey R. Moffitt, et al. Spatially resolved, highly multiplexed rna profiling in single cells. Science, 348(6233):aaa6090, 2015. 1

    work page 2015

  6. [6]

    Towards a general-purpose foundation model for computational pathol- ogy

    Richard J Chen, Tong Ding, Ming Y Lu, et al. Towards a general-purpose foundation model for computational pathol- ogy. Nature Medicine, 2024. 1

    work page 2024

  7. [7]

    Accurate spatial gene expression prediction by integrating multi-resolution features

    Youngmin Chung, Ji Hun Ha, Kyeong Chan Im, and Joo Sang Lee. Accurate spatial gene expression prediction by integrating multi-resolution features. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 11591–11600, 2024. 1, 2, 3, 5

    work page 2024

  8. [8]

    Borm, Amit Zeisel, et al

    Simone Codeluppi, Lars E. Borm, Amit Zeisel, et al. Spatial organization of the somatosensory cortex revealed by cyclic smfish. bioRxiv, 2018. 1

    work page 2018

Show all 43 references
  1. [9]

    Sys- tematic comparison of single-cell and single-nucleus rna- sequencing methods

    Jiarui Ding, Xian Adiconis, Sean K Simmons, et al. Sys- tematic comparison of single-cell and single-nucleus rna- sequencing methods. Nature biotechnology, 38(6):737–746,

  2. [10]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy. An image is worth 16x16 words: Transformers for image recognition at scale. arXiv preprint arXiv:2010.11929, 2020. 2, 3

    2010 arXiv

  3. [11]

    Transcriptome-scale super-resolved imaging in tissues by rna seqfish+

    Chee-Huat Linus Eng, Michael Lawson, Qian Zhu, et al. Transcriptome-scale super-resolved imaging in tissues by rna seqfish+. Nature, 568(7751):235–239, 2019. 1

    2019

  4. [12]

    Inductive representation learning on large graphs

    Will Hamilton, Zhitao Ying, and Jure Leskovec. Inductive representation learning on large graphs. Advances in neural information processing systems, 30, 2017. 3

    2017

  5. [13]

    In- tegrating spatial gene expression and breast tumour morphol- ogy via deep learning

    Bryan He, Ludvig Bergenstr ˚ahle, Linnea Stenbeck, et al. In- tegrating spatial gene expression and breast tumour morphol- ogy via deep learning. Nature biomedical engineering, 4(8): 827–834, 2020. 1, 2, 5, 6, 7, 3

    2020

  6. [14]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceed- ings of the IEEE conference on computer vision and pattern recognition, pages 770–778, 2016. 3, 5

    2016

  7. [15]

    A systematic evaluation of single-cell rna-sequencing imputation methods

    Wenpin Hou, Zhicheng Ji, Hongkai Ji, and Stephanie C Hicks. A systematic evaluation of single-cell rna-sequencing imputation methods. Genome biology, 21:1–30, 2020. 4

    2020

  8. [16]

    Densely connected convolutional net- works

    Gao Huang, Zhuang Liu, Laurens Van Der Maaten, and Kil- ian Q Weinberger. Densely connected convolutional net- works. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4700–4708, 2017. 2

    2017

  9. [17]

    Saver: gene expression recovery for single-cell rna sequencing

    Mo Huang, Jingshu Wang, Eduardo Torre, et al. Saver: gene expression recovery for single-cell rna sequencing. Nature methods, 15(7):539–542, 2018. 4

    2018

  10. [18]

    A visual–language foundation model for pathology image anal- ysis using medical twitter

    Zhi Huang, Federico Bianchi, Mert Yuksekgonul, et al. A visual–language foundation model for pathology image anal- ysis using medical twitter. Nature medicine, 29(9):2307– 2316, 2023. 1

    2023

  11. [19]

    Mrna stability and overexpression of fatty acid synthase in human breast cancer cell lines

    Dirk A Hunt, Hilary M Lane, Matthew E Zygmont, et al. Mrna stability and overexpression of fatty acid synthase in human breast cancer cell lines. Anticancer research, 27(1A): 27–34, 2007. 7

    2007

  12. [20]

    Hest- 1k: A dataset for spatial transcriptomics and histology image analysis

    Guillaume Jaume, Paul Doucet, Andrew H Song, et al. Hest- 1k: A dataset for spatial transcriptomics and histology image analysis. arXiv preprint arXiv:2406.16192, 2024. 2

    2024 arXiv

  13. [21]

    Multimodal analysis of composition and spatial architecture in human squamous cell carcinoma

    Andrew L Ji, Adam J Rubin, Kim Thrane, et al. Multimodal analysis of composition and spatial architecture in human squamous cell carcinoma. Cell, 182(2):497–514, 2020. 6, 7

    2020

  14. [22]

    Thitogene: a deep learning method for predicting spatial transcriptomics from histological images

    Yuran Jia, Junliang Liu, Li Chen, et al. Thitogene: a deep learning method for predicting spatial transcriptomics from histological images. Briefings in Bioinformatics, 25(1): bbad464, 2024. 7, 5

    2024

  15. [23]

    Elevated expression of gnas pro- motes breast cancer cell proliferation and migration via the pi3k/akt/snail1/e-cadherin axis

    X Jin, L Zhu, Z Cui, et al. Elevated expression of gnas pro- motes breast cancer cell proliferation and migration via the pi3k/akt/snail1/e-cadherin axis. Clinical and Translational Oncology, 21:1207–1219, 2019. 7

    2019

  16. [24]

    Bayesian approach to single-cell differential expression anal- ysis

    Peter V Kharchenko, Lev Silberstein, and David T Scadden. Bayesian approach to single-cell differential expression anal- ysis. Nature methods, 11(7):740–742, 2014. 2

    2014

  17. [25]

    Fast, sen- sitive and accurate integration of single-cell data with har- mony

    Ilya Korsunsky, Nghia Millard, Jean Fan, et al. Fast, sen- sitive and accurate integration of single-cell data with har- mony. Nature methods, 16(12):1289–1296, 2019. 3

    2019

  18. [26]

    SPCS: a spatial and pattern combined smoothing method for spatial transcriptomic expression

    Yusong Liu, Tongxin Wang, Ben Duggan, et al. SPCS: a spatial and pattern combined smoothing method for spatial transcriptomic expression. Briefings in Bioinformatics , 23 (3):bbac116, 2022. 2, 4, 6, 7, 1

    2022

  19. [27]

    Visual language pretrained multiple instance zero-shot transfer for histopathology images

    Ming Y Lu, Bowen Chen, Andrew Zhang, et al. Visual language pretrained multiple instance zero-shot transfer for histopathology images. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 19764–19775, 2023. 1

    2023

  20. [28]

    A visual-language foundation model for computational pathol- ogy

    Ming Y Lu, Bowen Chen, Drew FK Williamson, et al. A visual-language foundation model for computational pathol- ogy. Nature Medicine, 30(3):863–874, 2024. 1

    2024

  21. [29]

    EAGS: efficient and adaptive Gaussian smoothing applied to high-resolved spatial transcriptomics

    Tongxuan Lv, Ying Zhang, Mei Li, et al. EAGS: efficient and adaptive Gaussian smoothing applied to high-resolved spatial transcriptomics. GigaScience, 13:giad097, 2024. 4

    2024

  22. [30]

    Leverag- ing information in spatial transcriptomics to predict super- resolution gene expression from histology images in tumors

    Minxing Pang, Kenong Su, and Mingyao Li. Leverag- ing information in spatial transcriptomics to predict super- resolution gene expression from histology images in tumors. bioRxiv, 2021. 1, 2, 3, 5

    2021

  23. [31]

    St ˚ahl, Fredrik Salm ´en, Sanja Vickovic, et al

    Patrik L. St ˚ahl, Fredrik Salm ´en, Sanja Vickovic, et al. Vi- sualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science, 353(6294):78–82, 2016. 1, 2, 3, 5, 6

    2016

  24. [32]

    Patches are all you need? arXiv preprint arXiv:2201.09792, 2022

    Asher Trockman and J Zico Kolter. Patches are all you need? arXiv preprint arXiv:2201.09792, 2022. 3 9

    2022 arXiv

  25. [33]

    Re- covering gene interactions from single-cell data using data diffusion

    David Van Dijk, Roshan Sharma, Juozas Nainys, et al. Re- covering gene interactions from single-cell data using data diffusion. Cell, 174(3):716–729, 2018. 4

    2018

  26. [34]

    Attention is all you need

    A Vaswani. Attention is all you need. Advances in Neural Information Processing Systems, 2017. 5

    2017

  27. [35]

    Graph Attention Networks

    Petar Veli ˇckovi´c, Guillem Cucurull, Arantxa Casanova, et al. Graph Attention Networks. International Conference on Learning Representations, 2018. 4, 5

    2018

  28. [36]

    High-definition spatial transcriptomics for in situ tissue pro- filing

    Sanja Vickovic, G ¨okcen Eraslan, Fredrik Salm ´en, et al. High-definition spatial transcriptomics for in situ tissue pro- filing. Nature methods, 16(10):987–990, 2019. 1

    2019

  29. [37]

    A foundation model for clinical-grade computational pathol- ogy and rare cancers detection.Nature medicine, pages 1–12,

    Eugene V orontsov, Alican Bozkurt, Adam Casson, et al. A foundation model for clinical-grade computational pathol- ogy and rare cancers detection.Nature medicine, pages 1–12,

  30. [38]

    Allen, Matthew A

    Xiao Wang, William E. Allen, Matthew A. Wright, et al. Three-dimensional intact-tissue sequencing of single-cell transcriptional states. Science, 361(6400):eaat5691, 2018. 1

    2018

  31. [39]

    Spatially re- solved gene expression prediction from histology images via bi-modal contrastive learning

    Ronald Xie, Kuan Pang, Sai W Chung, et al. Spatially re- solved gene expression prediction from histology images via bi-modal contrastive learning. In Thirty-seventh Conference on Neural Information Processing Systems, 2023. 1, 2, 3

    2023

  32. [40]

    Exemplar guided deep neural network for spatial tran- scriptomics analysis of gene expression prediction

    Yan Yang, Md Zakir Hossain, Eric A Stone, and Shafin Rah- man. Exemplar guided deep neural network for spatial tran- scriptomics analysis of gene expression prediction. In Pro- ceedings of the IEEE/CVF Winter Conference on Applica- tions of Computer Vision, pages 5039–5048, 2023. 2, 3

    2023

  33. [41]

    Spatial tran- scriptomics prediction from histology jointly through Trans- former and graph neural networks

    Yuansong Zeng, Zhuoyi Wei, Weijiang Yu, et al. Spatial tran- scriptomics prediction from histology jointly through Trans- former and graph neural networks. Briefings in Bioinformat- ics, 23(5):bbac297, 2022. 1, 2, 3, 5 10 MERGE: Multi-faceted Hierarchical Graph-based GNN for G...

    2022

  34. [42]

    Experimental Setup The ResNet18-based patch encoder is implemented in Py- Torch (version 2.2.2)

    Implementation Details 7.1. Experimental Setup The ResNet18-based patch encoder is implemented in Py- Torch (version 2.2.2). The graph neural network is im- plemented using PyTorch Geometric (version 2.5.2). Both models are trained on a NVIDIA RTX A6000 GPU. To en- sure reprod...

  35. [43]

    Results Comparison - PCC This section discusses the PCC scores attained by MERGE and TRIPLEX across the samples in ST-Net dataset for two cancer-relevant genes - FASN and GNAS

    Results Analysis 8.1. Results Comparison - PCC This section discusses the PCC scores attained by MERGE and TRIPLEX across the samples in ST-Net dataset for two cancer-relevant genes - FASN and GNAS. Fig. 14 shows two bar charts depicting the PCC attained by MERGE and TRIPLEX f...

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.