Fig 1.
The flowchart included patients, samples and images.
A patient may have multiple tumor samples, and not all patients necessarily have normal samples available. Images showing obvious contamination, blurring, or blank areas exceeding 50% are defined as low-quality. Abbreviations: pts., patients; BrC., breast cancer.
Table 1.
Baseline Characteristics of Patients in the CHEK1-low and CHEK1-high Groups.
Fig 2.
Evaluation of the prognostic significance of CHEK1 in BrC using the TCGA-BRCA dataset.
A) The CHEK1 expression of BrC and normal breast tissues based on TCGA-BRCA RNA-seq data. B) The KM curves of OS rate based on CHEK1 expression in BrC tissues using 96 months post-diagnosis as the landmark. C) The univariate and multivariate Cox regression analyses of potential risk factors for OS. An HR value greater than 1 is a risk factor, while an HR value less than 1 is a protective factor. Abbreviations: BrC., breast cancer; *, P < 0.001; m, months; OS, overall survival; HR, hazard ratio.
Fig 3.
The workflow for generating, evaluating, exploring the implication, and validating the CHEK1 prediction pathomics model.
1) H&E-stained images in svs format, with a maximum magnification of 20 × or 40 × , are used. 2) 40 × images are divided into 1024 × 1024 sub-images, while 20 × images are magnified to 1024 × 1024 from 512 × 512 sub-images. 3) Images are manually reviewed by two independent pathologists. Those with obvious contamination, blurring, or blank areas exceeding 50% are defined as low quality. 4) The performance of the pathomics model is assessed using the R packages “pROC”, “ResourceSelection”, “rms”, and “rmda”. 5) The pre-defined Hallmark gene sets, and the KEGG gene sets are used for analysis. 6) Including immune cell infiltration, immune gene expression, and potential response to anti-PD-1 and anti-CTLA4 treatments are explored. 7) IHC is the abbreviation for immunohistochemistry.
Fig 4.
Performance evaluation of the CHEK1 prediction pathomics model.
A) The top 20 features from analyzed H&E-stained images are selected using the mRMR algorithm. B) The RFE algorithm further identifies the 8 most important features. The graph shows the ROC curves of our pathomics model in the C) training and F) validation sets. The graph shows the calibration curve of our pathomics model in the D) training and G) validation sets, along with the DCA curves of our pathomics model in the E) training and H) validation sets.
Table 2.
Baseline Characteristics of Patients in the PS-high and PS-low Groups.
Fig 5.
Evaluation of the prognostic significance of PS values in BrC using the TCGA-BRCA dataset.
The correlation between PS and CHEK1 expression levels in the A) training and B) validation sets are assessed along with the statistical differences using the Wilcoxon test. C) The graph shows the KM curves for OS in patients with high and low PS. The median OS for the “PS-high” and “PS-low” groups are indicated in red and blue text, respectively. D) The univariate and multivariate Cox regression analyses of potential risk factors for OS are presented. An HR value greater than 1 indicates a risk factor, while an HR value less than 1 indicates a protective factor. Abbreviations: *, P < 0.001; m, months; OS, overall survival; HR, hazard ratio.
Fig 6.
Potential implications of high PS in BrC.
The image shows the enriched pathways using A) the KEGG gene set and B) the Hallmark gene set from Gene Set Enrichment Analysis (GSEA). C) The graph shows the differences in the immune gene expression between the PS-high and PS-low groups. D) The data show the differences in immune cell infiltration between the PS-high and PS-low groups. E) The TIDE scores predict the differences in response to anti-PD-1 and anti-CTLA4 treatments between the PS-high and PS-low groups.
Fig 7.
Validation of the CHEK1 prediction pathomics model using BrC TMA.
A) The workflow validates the real-world performance of this CHEK1 prediction pathomics model using BrC TMA slides.B) The data show the H&E-stained image of the BrC in TMA.C) The IHC image of the BrC TMA shows the stained CHEK1 antibodies.
Fig 8.
Correlation and survival analysis of CHEK1 expression in BrC TMA.
A) The image shows the correlation between predicted PS and actual CHEK1 expression levels (Histochemistry Score, HS) in TMA, where statistical differences are assessed using the Wilcoxon test. E) The data from TMA present the KM curves for OS in patients with high and low PS.