CoChat

Seeing or believing in hyperplexed spatial proteomics via antibodies: New and old biases for an image-based technology

Open access

Published version of a preprint

The DOI matches its Crossref record; 71 of the 71 references it cites were found in a registry or an index.

Abstract

Hyperplexed in-situ targeted proteomics via antibody immunodetection (i.e., >15 markers) is changing how we classify cells and tissues. Differently from other high-dimensional single-cell assays (flow cytometry, single-cell RNA sequencing), the human eye is a necessary component in multiple procedural steps: image segmentation, signal thresholding, antibody validation, and iconographic rendering. Established methods complement the human image evaluation, but may carry undisclosed biases in such a new context, therefore we re-evaluate all the steps in hyperplexed proteomics. We found that the human eye can discriminate less than 64 out of 256 gray levels and has limitations in discriminating luminance levels in conventional histology images. Furthermore, only images containing visible signals are selected and eye-guided digital thresholding separates signal from noise. BRAQUE, a hyperplexed proteomic tool, can extract, in a marker-agnostic fashion, granular information from markers which have a very low signal-to-noise ratio and therefore are not visualized by traditional visual rendering. By analyzing a public human lymph node dataset, we also found unpredicted staining results by validated antibodies, which highlight the need to upgrade the definition of antibody specificity in hyperplexed immunostaining. Spatially hyperplexed methods upgrade and supplant traditional image-based analysis of tissue immunostaining, beyond the human eye contribution.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1017/s2633903x24000138

Cites

71 of 71 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1. (a) The shades of gray score distribution of 15 Pathologists, ordered progressively. The number inside each circle is the number of subjects with that score. (b) The relationship between the score (x-axis) and the subject’s age (y-axis). The R-squared value of the intercept is R2 = 0.0856. (c) The relationship between the score (x-axis) and the pathologist’s working experience in years (y-axis). The R-squared value of the intercept is R2 = 0.1876.
Figure 1. (a) The shades of gray score distribution of 15 Pathologists, ordered progressively. The number inside each circle is the number of subjects with that score. (b) The relationship between the score (x-axis) and the subject’s age (y-axis). The R-squared value of the intercept is R2 = 0.0856. (c) The relationship between the score (x-axis) and the pathologist’s working experience in years (y-axis). The R-squared value of the intercept is R2 = 0.1876.CC BY · PubMed Central
Figure 2. Example of two different bit depths for an H&E image. The two images represent A, a high magnification detail of an 8 bit (24 bits in RGB) H&E image of the human colon, B the same image at 4 bit depth reduction. The R, G, and B images on the right are the frequency plots of the image pixels, distributed along the 0–255 channels for each of the three color components. Image A contains 255 levels per channel, image B 16 levels, as shown by the laddering of the RGB profile in the RGB details. Note the same size of the visible pixels. Scale bar: 2.25 μm (five pixels; 3200x). The insets in the lower left corner of each image show the full-size originating images (scale bar 50 μm).
Figure 2. Example of two different bit depths for an H&E image. The two images represent A, a high magnification detail of an 8 bit (24 bits in RGB) H&E image of the human colon, B the same image at 4 bit depth reduction. The R, G, and B images on the right are the frequency plots of the image pixels, distributed along the 0–255 channels for each of the three color components. Image A contains 255 levels per channel, image B 16 levels, as shown by the laddering of the RGB profile in the RGB details. Note the same size of the visible pixels. Scale bar: 2.25 μm (five pixels; 3200x). The insets in the lower left corner of each image show the full-size originating images (scale bar 50 μm).CC BY · PubMed Central
Figure 3. Descriptive graphics of the bit-reduced images scores. (a) The two-scale image shows the bit depth below which each of the 12 pathologists identifies degradation on a monochrome image (red circles; scale to the right). The mean ± SD percentage of correct bit depth identification on the global test is shown (blue squares; scale to the left). (b) Mean percentage ± SD of correct identification on images divided by bit depth. (c) Mean percentage ± SD of correct identification on images divided by image type. B&W: grayscale images; Color: H&E-stained images; IHC: immunohistochemistry examples; Special: special stains; IF: immunofluorescence. (d) Mean percentage ± SD of correct bit depth identification subdivided between top discrimination (discrimination between 8 and 6 bits), bottom discrimination (5 and 4 bits), 8 bits versus 4 bits, 5 or 6 bits versus 4 bits, 8 bits versus 5 or 6 bits. The scores are further shown for the whole test or divided by image type (B&W: grayscale images; Color: H&E-stained images; IHC: immunohistochemistry examples; Special: special stains).
Figure 3. Descriptive graphics of the bit-reduced images scores. (a) The two-scale image shows the bit depth below which each of the 12 pathologists identifies degradation on a monochrome image (red circles; scale to the right). The mean ± SD percentage of correct bit depth identification on the global test is shown (blue squares; scale to the left). (b) Mean percentage ± SD of correct identification on images divided by bit depth. (c) Mean percentage ± SD of correct identification on images divided by image type. B&W: grayscale images; Color: H&E-stained images; IHC: immunohistochemistry examples; Special: special stains; IF: immunofluorescence. (d) Mean percentage ± SD of correct bit depth identification subdivided between top discrimination (discrimination between 8 and 6 bits), bottom discrimination (5 and 4 bits), 8 bits versus 4 bits, 5 or 6 bits versus 4 bits, 8 bits versus 5 or 6 bits. The scores are further shown for the whole test or divided by image type (B&W: grayscale images; Color: H&E-stained images; IHC: immunohistochemistry examples; Special: special stains).CC BY · PubMed Central
Figure 4. Comparison of sensitivity of detection systems. The area of detection of anti-LMW KRT in serial LN section by secondary Abs conjugated with three fluorochromes (FITC = Alexa 488 green, TRITC = Rhodamine Red™ X orange, Cy5 = Alexa 647 red) and DAB is plotted on a logarithmic scale, relative to the area detected with TSA Alexa 647 (100%). Duplicate experiments.
Figure 4. Comparison of sensitivity of detection systems. The area of detection of anti-LMW KRT in serial LN section by secondary Abs conjugated with three fluorochromes (FITC = Alexa 488 green, TRITC = Rhodamine Red™ X orange, Cy5 = Alexa 647 red) and DAB is plotted on a logarithmic scale, relative to the area detected with TSA Alexa 647 (100%). Duplicate experiments.CC BY · PubMed Central
Figure 5. Thresholding LMW KRT staining in LN. The LMW KRT IF stain detail is shown as an inverted, unmodified image, modified with three different Fiji thresholding algorithms (Otsu, Huang, and percentile) and as a 3D plot. Only three thresholding algorithms are shown, out of 17 tested. A FRC CK+ dendrite is highlighted with a red rectangle and the mean pixel density value along that rectangle is shown next to the image. Note that, because of the image inversion, darker pixels have lower values in the plot. The continuous intensity variations of the signal above background can be appreciated in the 3D plots. Note that the percentile algorithm highlights numerous background spots in addition to the dendrite of interest. The image shown measures 105 × 165 pixels (47.5 × 74.25 μm).
Figure 5. Thresholding LMW KRT staining in LN. The LMW KRT IF stain detail is shown as an inverted, unmodified image, modified with three different Fiji thresholding algorithms (Otsu, Huang, and percentile) and as a 3D plot. Only three thresholding algorithms are shown, out of 17 tested. A FRC CK+ dendrite is highlighted with a red rectangle and the mean pixel density value along that rectangle is shown next to the image. Note that, because of the image inversion, darker pixels have lower values in the plot. The continuous intensity variations of the signal above background can be appreciated in the 3D plots. Note that the percentile algorithm highlights numerous background spots in addition to the dendrite of interest. The image shown measures 105 × 165 pixels (47.5 × 74.25 μm).CC BY · PubMed Central
Figure 6. Tissue distribution of FRC and stromal cells. The distribution of CK+ FRC (red; clusters 3, 6, 12, 22, 23, 56, 60, 78, 87), fibroblasts & myofibroblasts (blue; clusters 24, 31, 32, 36, 54, 59, 73, 82) and endothelial cells (green; clusters 13, 14, 17, 25, 28, 29, 37, 46, 53, 72) is shown plotted on the UMAP plot (left) and on a gray image of the LN section (right). The x and y scales on the left are UMAP virtual space arbitrary references, on the right real pixel image dimensions (0.45 μm per pixel). The gray outlines represent the remaining cell clusters (left) and the total of single cells (right).
Figure 6. Tissue distribution of FRC and stromal cells. The distribution of CK+ FRC (red; clusters 3, 6, 12, 22, 23, 56, 60, 78, 87), fibroblasts & myofibroblasts (blue; clusters 24, 31, 32, 36, 54, 59, 73, 82) and endothelial cells (green; clusters 13, 14, 17, 25, 28, 29, 37, 46, 53, 72) is shown plotted on the UMAP plot (left) and on a gray image of the LN section (right). The x and y scales on the left are UMAP virtual space arbitrary references, on the right real pixel image dimensions (0.45 μm per pixel). The gray outlines represent the remaining cell clusters (left) and the total of single cells (right).CC BY · PubMed Central
Figure 7. Image rendering of CK+ FRC. (a) A detail of three markers, CK AE1–AE3, Vimentin and SMA as 2D log-transformed and inverted fluorescence CODEX images (top row) and the 3D transformation of each (bottom row), is shown. (b) The same type of images are shown for another tissue detail, comparing CK AE1-AE3 and a B cell marker, CD20. Scale bar: 100 μm. (c) A low and a high power magnification detail show a three-color image of a LN, produced in-house, where nuclei are blue (DAPI) CK AE3 is green and a LMW KRT cocktail is red. In this image, note a B cell follicle (asterisk) and coexpression of AE3 and LMW KRT in FRC as yellow (green + red). Scale bar: 100 μm.
Figure 7. Image rendering of CK+ FRC. (a) A detail of three markers, CK AE1–AE3, Vimentin and SMA as 2D log-transformed and inverted fluorescence CODEX images (top row) and the 3D transformation of each (bottom row), is shown. (b) The same type of images are shown for another tissue detail, comparing CK AE1-AE3 and a B cell marker, CD20. Scale bar: 100 μm. (c) A low and a high power magnification detail show a three-color image of a LN, produced in-house, where nuclei are blue (DAPI) CK AE3 is green and a LMW KRT cocktail is red. In this image, note a B cell follicle (asterisk) and coexpression of AE3 and LMW KRT in FRC as yellow (green + red). Scale bar: 100 μm.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 71 references

Cited by

See all 3

Metadata

Authors
Bolognesi, M. M., Dall’Olio, L., Maerten, A., Borghesi, S., Castellani, G., & Cattoretti, G.
Affiliations
  • Institute of Molecular Bioimaging and Physiology (IT)
  • Istituto delle Scienze Neurologiche di Bologna (IT)
  • Vrije Universiteit Brussel (BE)
  • University of Milano-Bicocca (IT)
  • University of Bologna (IT)
Corresponding author
Cattoretti, Giorgio
Venue
Biological Imaging
Published
2024
Volume
4
Pages
e10-e10
ISSN
2633-903X (ISSN-L)
OpenAlex
W4403694216
Cited by
3
References
70
Citations by year
2026: 1 · 2025: 2
FWCI
0.41
Type
journal article
Language
en
Open access
Locations
  • Biological Imaging (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • BOA (University of Milano-Bicocca) (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • Associazione Italiana per la Ricerca sul Cancro
  • Regione Lombardia (2014–2020)
  • European Regional Development Fund (2014-2020’)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Single-cell and spatial transcriptomics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Biophysics › Cell Image Analysis Techniques
  • Physical Sciences › Chemistry › Spectroscopy › Advanced Proteomics Techniques and Applications
Keywords
visual perception, image segmentation, signal thresholding, antibody validation, antibody specificity, Braque, signal-to-noise ratio, lymph nodes, histological image analysis, visual bias

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Seeing or believing in hyperplexed spatial proteomics via antibodies: New and old biases for an image-based technology. https://app.cochat.ai/papers/10.1017/s2633903x24000138
The paper
Bolognesi, M. M., Dall’Olio, L., Maerten, A., Borghesi, S., Castellani, G., & Cattoretti, G. (2024). Seeing or believing in hyperplexed spatial proteomics via antibodies: New and old biases for an image-based technology. Biological Imaging. https://doi.org/10.1017/s2633903x24000138

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.