Title?

題目

Spatial cellular architecture predicts prognosis in glioblastoma

空間細(xì)胞結(jié)構(gòu)預(yù)測(cè)膠質(zhì)母細(xì)胞瘤的預(yù)后

01文獻(xiàn)速遞介紹

膠質(zhì)母細(xì)胞瘤的治療耐藥性的關(guān)鍵驅(qū)動(dòng)因素是腫瘤內(nèi)的異質(zhì)性和細(xì)胞狀態(tài)的可塑性。在這里，我們調(diào)查了空間細(xì)胞組織與膠質(zhì)母細(xì)胞瘤預(yù)后之間的關(guān)聯(lián)。利用單細(xì)胞RNA測(cè)序和空間轉(zhuǎn)錄組數(shù)據(jù)，我們開發(fā)了一個(gè)深度學(xué)習(xí)模型，從組織學(xué)圖像中預(yù)測(cè)膠質(zhì)母細(xì)胞瘤細(xì)胞的轉(zhuǎn)錄亞型。利用這個(gè)模型，我們對(duì)來自410名患者的4000萬個(gè)組織點(diǎn)進(jìn)行了表型分析，并在兩個(gè)獨(dú)立的隊(duì)列中確定了腫瘤結(jié)構(gòu)與預(yù)后之間的一致關(guān)聯(lián)。預(yù)后不佳的患者表現(xiàn)出更高比例的腫瘤細(xì)胞表達(dá)缺氧誘導(dǎo)的轉(zhuǎn)錄程序。此外，類似星形膠質(zhì)細(xì)胞的腫瘤細(xì)胞的聚類模式與預(yù)后不佳相關(guān)，而星形膠質(zhì)細(xì)胞與其他轉(zhuǎn)錄亞型的分散和連接與風(fēng)險(xiǎn)減少相關(guān)。為了驗(yàn)證這些結(jié)果，我們開發(fā)了一個(gè)單獨(dú)的深度學(xué)習(xí)模型，利用組織學(xué)圖像預(yù)測(cè)預(yù)后。將這個(gè)模型應(yīng)用于空間轉(zhuǎn)錄組數(shù)據(jù)，揭示了與生存相關(guān)的區(qū)域基因表達(dá)程序?？偟膩碚f，我們的研究提出了一種可擴(kuò)展的方法來揭示膠質(zhì)母細(xì)胞瘤的轉(zhuǎn)錄異質(zhì)性，并建立了空間細(xì)胞結(jié)構(gòu)與臨床結(jié)果之間的重要聯(lián)系。

Results

結(jié)果

Identifications of spatially resolved transcriptional subtypes To resolve the transcriptional heterogeneity of GBM within the spatial context, we performed an integrative analysis of three spatial tran?scriptomics datasets (Supplementary Data 1)16,18. The integrated data?set comprised 23 GBM samples obtained from 22 patients. Each?sample contains 2500 ~4702 gene expression spots, resulting in 75,625transcriptomes. Data preprocessing and batch-effect normalization?were described in the “Methods” section. To determine the number of?cells in each spot, we performed nuclei segmentation on histology?images. The cell count ranged from 3 to 38, with an average count of 13?cells (Supplementary Fig. 1a). To determine genomic abnormalities of?the GBM samples, we inferred copy number alterations (CNAs) using?the transcriptomics profile of each spot, where data from a separate?cohort of normal brain tissues (n = 6 tissues from 3 patients) were used?as a reference31. Tumor samples demonstrated broad CNAs across?chromosomes, including gains of Chr 6, Chr 7 and loss of Chr 8, Chr 10?and Chr14 (Fig. 1a and Supplementary Fig. 1b). Since GBM cells are?highly infiltrative, each spot may contain a mixture of tumor cells and?normal brain tissues. To estimate the tumor cell content within each?spot, we first designated a prominent CNA event that shared across all?the spots in each tumor as tumor signature CNA. The tumor cell con?tent was then estimated based on the score of the CNA signature?(“Methods”). At least three signature events were calculated in each?tumor to ensure robust and unbiased estimations. We found that our?approach was able to distinguish tumor regions versus histologically?normal peripheral tissues (Supplementary Figs. 1c, d). Therefore, we?used CNA-based estimation of tumor cell contents to filter malignant?spots, while spots with low (<20%) tumor cell content were removed in?our subsequent analysis.

為了解析膠質(zhì)母細(xì)胞瘤（GBM）在空間上的轉(zhuǎn)錄異質(zhì)性，我們對(duì)三個(gè)空間轉(zhuǎn)錄組數(shù)據(jù)集進(jìn)行了整合分析（補(bǔ)充資料1）16,18。整合數(shù)據(jù)集包括來自22位患者的23個(gè)GBM樣本。每個(gè)樣本包含2500~4702個(gè)基因表達(dá)點(diǎn)，共計(jì)75,625個(gè)轉(zhuǎn)錄組。數(shù)據(jù)預(yù)處理和批次效應(yīng)標(biāo)準(zhǔn)化在“方法”部分有描述。為了確定每個(gè)點(diǎn)中的細(xì)胞數(shù)量，我們?cè)诮M織學(xué)圖像上進(jìn)行了核分割。細(xì)胞計(jì)數(shù)范圍從3到38，平均計(jì)數(shù)為13個(gè)細(xì)胞（補(bǔ)充圖1a）。為了確定GBM樣本的基因組異常，我們利用每個(gè)點(diǎn)的轉(zhuǎn)錄組概況推斷拷貝數(shù)變異（CNA），其中使用來自3位患者的6份正常腦組織（n = 6 tissues from 3 patients）作為參考31。腫瘤樣本顯示出染色體廣泛的CNA，包括第6、第7染色體的增加和第8、第10及第14染色體的丟失（圖1a及補(bǔ)充圖1b）。由于GBM細(xì)胞具有高度浸潤性，每個(gè)點(diǎn)可能包含腫瘤細(xì)胞和正常腦組織的混合物。為了估算每個(gè)點(diǎn)中的腫瘤細(xì)胞含量，我們首先指定每個(gè)腫瘤中所有點(diǎn)共有的一個(gè)顯著CNA事件作為腫瘤特征CNA。然后根據(jù)CNA特征的得分估算腫瘤細(xì)胞含量（“方法”）。每個(gè)腫瘤至少計(jì)算三個(gè)特征事件，以確保估算的健壯性和無偏性。我們發(fā)現(xiàn)這種方法能夠區(qū)分腫瘤區(qū)域和組織學(xué)正常的外圍組織（補(bǔ)充圖1c, d）。因此，我們使用基于CNA的腫瘤細(xì)胞含量估算方法來篩選惡性點(diǎn)，而腫瘤細(xì)胞含量低于20%的點(diǎn)在后續(xù)分析中被移除。

Method

方法

Preprocessing of the spatial transcriptomics dataWe used four publicly available spatial transcriptomics datasets comprising both tumors and normal brain tissues (SupplementaryData 1)16,18. All datasets were generated using the 10X Visium platform.Quality control was performed by the cell ranger pipeline and imported into AnnData objects using the Scanpy software (version 1.9). Ineach sample, we removed spots with less than 200 detected genes andmore than 5% mitochondrial RNA. Additionally, genes detected in lessthan 3 spots were removed. Given the potential presence of batcheffects in spatial transcriptomics data, we performed normalizationand variance stabilization across different samples using regularizednegative binomial regression55. We regressed out percentages ofmitochondria-expressed genes per spot and effects from cell cycles.This approach allowed us to remove the influence of technical variances from downstream analyses while preserving biologicalheterogeneity.

空間轉(zhuǎn)錄組數(shù)據(jù)的預(yù)處理 我們使用了四個(gè)公開可用的空間轉(zhuǎn)錄組數(shù)據(jù)集，包括腫瘤和正常腦組織（補(bǔ)充數(shù)據(jù) 1）16,18。所有數(shù)據(jù)集均使用 10X Visium 平臺(tái)生成。通過 cell ranger 流程進(jìn)行質(zhì)量控制，并使用 Scanpy 軟件（版本 1.9）導(dǎo)入到 AnnData 對(duì)象中。在每個(gè)樣本中，我們移除了檢測(cè)到的基因少于 200 個(gè)且線粒體 RNA 超過 5% 的斑點(diǎn)。此外，還移除了在少于 3 個(gè)斑點(diǎn)中檢測(cè)到的基因。鑒于空間轉(zhuǎn)錄組數(shù)據(jù)中可能存在的批次效應(yīng)，我們使用規(guī)范化的負(fù)二項(xiàng)回歸進(jìn)行了歸一化和方差穩(wěn)定處理。我們回歸了每個(gè)斑點(diǎn)中線粒體表達(dá)基因的百分比和細(xì)胞周期的影響。這種方法使我們能夠在保留生物學(xué)異質(zhì)性的同時(shí)，去除技術(shù)變異對(duì)下游分析的影響。

Figure

圖

Fig. 1 | Identifications of spatial gene expression programs in GBM. a Heatmaps?showing the tumor cell content across different spots and corresponding CNAs?across different chromosomes in a representative sample. b Heatmap showing?gene expression levels of the top 60 signatures from each cNMF module. Malignant?spots (n = 69, 647) from all samples (n = 23) were grouped by the expression score?of each module. c, d Heatmaps showing the average correlation coefficients?(n = 23 samples) from spatially weighted correlation analysis between the cNMF?modules (x-axis) and published modules from (c) Ravi et al. 16 and (d) Neftel et al. 5Two-sided Wald tests were used to determine statistical significance, and P values?were adjusted for multiple testing using the Benjamini-Hochberg procedure.P < 0.05, P < 0.01, *P < 0.001. e Stacked bar plot showing the fractions of different transcriptional subtypes in each sample. Transcriptional subtype was?determined using the top-scoring cNMF module in each spot. f spatial visualizationshowing the distribution of transcriptional subtypes in two example tumors. Spots?were colored by transcriptional subtypes as indicated in panel e. g Pipeline for?computational deconvolution of spots using single-cell RNA-seq data as reference:I. UMAP visualization of the reference single-cell RNA-seq data. Each dot represents?a cell colored by the subtype; II. cell count estimation of each spot based on nuclei?segmentation; III. Align cell types from the reference dataset to spots. h Histogram?showing the fraction of dominant tumor cell type over all tumor cells in each spot(total n = 69, 647 spots; n = 23 samples). i The number of immune cell types in each?spot (total n = 69, 647 spots; n = 23 samples). j The average fraction (x-axis) of each?individual cell type from the single-cell RNA-seq data in spots classified by cNMFmodules (y-axis). Source data are provided as a Source Data file.

圖1 | 膠質(zhì)母細(xì)胞瘤中空間基因表達(dá)程序的識(shí)別。a 熱圖顯示不同點(diǎn)中的腫瘤細(xì)胞含量及相應(yīng)染色體上的拷貝數(shù)變異（CNA）。b 熱圖顯示每個(gè)cNMF模塊前60個(gè)標(biāo)志性基因的表達(dá)水平。所有樣本（n=23）中的惡性點(diǎn)（n=69,647）根據(jù)每個(gè)模塊的表達(dá)得分進(jìn)行分組。c, d 熱圖顯示空間加權(quán)相關(guān)分析中的平均相關(guān)系數(shù)（n=23個(gè)樣本）從cNMF模塊（x軸）到（c）Ravi等人16和（d）Neftel等人5發(fā)布的模塊。使用雙側(cè)Wald測(cè)試來確定統(tǒng)計(jì)顯著性，P值使用Benjamini-Hochberg程序進(jìn)行多重檢驗(yàn)調(diào)整。*P < 0.05, P < 0.01, *P < 0.001。e 堆疊條形圖顯示每個(gè)樣本中不同轉(zhuǎn)錄亞型的比例。轉(zhuǎn)錄亞型是根據(jù)每個(gè)點(diǎn)中得分最高的cNMF模塊確定的。f 空間可視化顯示兩個(gè)例子腫瘤中轉(zhuǎn)錄亞型的分布。點(diǎn)按照e面板所示的轉(zhuǎn)錄亞型進(jìn)行著色。g 使用單細(xì)胞RNA測(cè)序數(shù)據(jù)作為參考的點(diǎn)計(jì)算解卷積的流程圖：I. 參考單細(xì)胞RNA測(cè)序數(shù)據(jù)的UMAP可視化。每個(gè)點(diǎn)代表一種按亞型著色的細(xì)胞；II. 基于核分割估算每個(gè)點(diǎn)的細(xì)胞計(jì)數(shù)；III. 將參考數(shù)據(jù)集中的細(xì)胞類型與點(diǎn)對(duì)齊。h 直方圖顯示每個(gè)點(diǎn)中占主導(dǎo)地位的腫瘤細(xì)胞類型的比例（總n = 69,647點(diǎn)；n = 23個(gè)樣本）。i 每個(gè)點(diǎn)中免疫細(xì)胞類型的數(shù)量（總n = 69,647點(diǎn)；n = 23個(gè)樣本）。j 平均比例（x軸）顯示每個(gè)單細(xì)胞RNA測(cè)序數(shù)據(jù)中的單一細(xì)胞類型在由cNMF模塊分類的點(diǎn)中的分布（y軸）。源數(shù)據(jù)作為源數(shù)據(jù)文件提供。

Fig. 2 | Development and validation of GBM-CNN for spatially resolved tran?scriptional subtype prediction. a Architecture of GBM-CNN. Histology imageswere cropped to extract patches corresponding to each spot. Each patch was thentransformed into a feature vector (2048 × 1) using a ResNet-50 module. Subse?quently, each feature vector was mapped to a probability vector (8 × 1) through a?fully connected layer. The cell-type cartoons were created with BioRender.com.b Confusion matrix showing the classification performance of GBM-CNN in predicting the dominant tumor cell type. Predictions from all folds (n = 23) were?averaged into a single matrix. c, d Confusion matrices showing the classification?performance of GBM-CNN in predicting the presence of (c) T cell and (d) macro?phage. e, f Alignment of ground truth gene expression signals obtained from in situRNA hybridization and the predicted probability scores of individual cell types in?matched histological sections. Examples from two different tumors were presented. g, h H&E images and the predicted distribution of transcriptional subtypes?in two tumors from the TCGA cohort. Bar graphs depict transcriptional subtype?proportions derived from the image prediction versus bulk RNA-seq deconvolu?tion. i Heatmap of Pearson correlation coefficient showing the agreement between?transcriptional subtype proportions derived from our image predictions versus?those estimated from bulk RNA-seq deconvolution (n = 166 patients). P values were?determined using the two-sided Pearson correlation test and were adjusted by theBenjamini-Hochberg procedure. P < 0.01, *P < 0.001. Source data are providedas a Source Data file.

圖 2 | GBM-CNN 用于空間解析轉(zhuǎn)錄亞型預(yù)測(cè)的開發(fā)和驗(yàn)證。a GBM-CNN 架構(gòu)。利用組織學(xué)圖像截取每個(gè)斑點(diǎn)對(duì)應(yīng)的片段。然后，每個(gè)片段通過 ResNet-50 模塊轉(zhuǎn)換為特征向量（2048 × 1）。隨后，每個(gè)特征向量通過一個(gè)全連接層映射成概率向量（8 × 1）。細(xì)胞類型卡通圖由 BioRender.com 創(chuàng)建。b 混淆矩陣顯示 GBM-CNN 在預(yù)測(cè)主導(dǎo)腫瘤細(xì)胞類型方面的分類性能。所有折疊（n = 23）的預(yù)測(cè)平均成一個(gè)單獨(dú)的矩陣。c，d 混淆矩陣顯示 GBM-CNN 在預(yù)測(cè)（c）T細(xì)胞和（d）巨噬細(xì)胞存在方面的分類性能。e，f 從原位 RNA 雜交獲得的基因表達(dá)信號(hào)的真實(shí)情況與在匹配的組織學(xué)切片中預(yù)測(cè)的各個(gè)細(xì)胞類型的概率分?jǐn)?shù)的對(duì)齊。展示了來自兩個(gè)不同腫瘤的例子。g，h 來自 TCGA 隊(duì)列的兩個(gè)腫瘤的 H&E 圖像和預(yù)測(cè)的轉(zhuǎn)錄亞型分布。條形圖顯示了從圖像預(yù)測(cè)中得到的轉(zhuǎn)錄亞型比例與通過批量 RNA 測(cè)序解析得到的轉(zhuǎn)錄亞型比例。i 熱圖顯示了從我們的圖像預(yù)測(cè)得到的轉(zhuǎn)錄亞型比例與通過批量 RNA 測(cè)序解析估計(jì)的比例之間的皮爾遜相關(guān)系數(shù)的一致性（n = 166 名患者）。P 值通過雙側(cè)皮爾遜相關(guān)檢驗(yàn)確定，并通過 Benjamini-Hochberg 程序調(diào)整。P < 0.01，*P < 0.001。來源數(shù)據(jù)以來源數(shù)據(jù)文件形式提供。

Fig. 3 | Associations between spatial cellular architecture and prognosis.a Schematic representation of a spatial neighborhood graph. Each patch represents?a node and connections between patches are edges. b, c Hazard ratio (HR) for?frequency of transcriptional subtype interactions and prognosis using data from?the (b) TCGA (n = 312 patients) and (c) CPTAC (n = 98 patients) cohorts. Statistical?significance was determined using multivariate Cox regression analysis, and sig?nificant associations were highlighted by red for HR > 0 and blue for HR < 0.*P < 0.05, P < 0.01, *P < 0.001. d, e Representative tumor samples with high?clustering coefficient (CC) of the AC-like subtype. Spots were colored by tran?scriptional subtypes. Abstractive networks demonstrate tumor regions character?ized by clusters of AC-like spots, as indicated by white arrows. f, g Representative?tumor samples with low clustering coefficient (CC) of the AC-like subtype. Spots?were colored by transcriptional subtype. Abstractive networks highlighted the?interactions between the AC-like subtype and other subtypes, such as NPC-like,MES-like and MES-hypoxia. h Representative tumor sample with a high frequencyof interaction between the OPC-like and MES-hypoxia subtype. i Kaplan-Meiersurvival curves of TCGA patients with high (n = 156) and low (n = 156) interactions?between the OPC-like and MES-hypoxia subtypes. Error bands represent con-fidence intervals for the estimated survival probabilities, and the survival curves arecompared with the log-rank test (P = 0.01). Source data are providedas a SourceData file.

圖 3 | 空間細(xì)胞結(jié)構(gòu)與預(yù)后之間的關(guān)聯(lián)。a 空間鄰域圖的示意圖。每個(gè)片段代表一個(gè)節(jié)點(diǎn)，片段之間的連接是邊。b，c 轉(zhuǎn)錄亞型相互作用頻率與預(yù)后的危險(xiǎn)比（HR），使用來自 (b) TCGA（n = 312名患者）和 (c) CPTAC（n = 98名患者）隊(duì)列的數(shù)據(jù)。統(tǒng)計(jì)顯著性通過多變量 Cox 回歸分析確定，顯著關(guān)聯(lián)以紅色突出顯示 HR > 0，以藍(lán)色突出顯示 HR < 0。P < 0.05, **P < 0.01, P < 0.001。d，e 代表具有高聚類系數(shù)（CC）的 AC-like 亞型的腫瘤樣本。斑點(diǎn)按轉(zhuǎn)錄亞型著色。抽象網(wǎng)絡(luò)顯示由 AC-like 斑點(diǎn)簇組成的腫瘤區(qū)域，如白色箭頭所示。f，g 代表具有低聚類系數(shù)（CC）的 AC-like 亞型的腫瘤樣本。斑點(diǎn)按轉(zhuǎn)錄亞型著色。抽象網(wǎng)絡(luò)突出顯示了 AC-like 亞型與其他亞型（如 NPC-like、MES-like 和 MES-hypoxia）之間的互動(dòng)。h 代表高頻互動(dòng)的腫瘤樣本，涉及 OPC-like 與 MES-hypoxia 亞型。i Kaplan-Meier 生存曲線顯示 TCGA 患者中與 OPC-like 和 MES-hypoxia 亞型之間的高（n = 156）與低（n = 156）互動(dòng)。誤差帶代表估計(jì)生存概率的置信區(qū)間，生存曲線通過對(duì)數(shù)秩檢驗(yàn)（P = 0.01）進(jìn)行比較。來源數(shù)據(jù)以來源數(shù)據(jù)文件形式提供。

Fig. 4 | In situ identifications of gene expression markers associated with?prognosis. a A deep-learning model was trained on whole slide images from the?TCGA cohort to predict patient prognosis. H&E-stained histology images were?cropped into 56μm × 56μm patches. Each patch was converted to a feature?vector (2048 × 1) using a ResNet-50 module. The feature vectors were then?mapped to an aggressive score through a Cox regression module. The aggressive?scores of each patient were averaged for validation. b Using the trained image?model from panel (a) to predict aggressive scores for spots in spatial tran?scriptomics. c, d Visualization of transcriptional subtypes and the predicted?aggressive scores in two tumors from the spatial transcriptomics cohort.Aggressive scores were normalized within each sample using min-max normal?ization. e Bar plot of median aggressive scores for malignant spots. Aggressive?scores from all tumors (n = 23) were pooled together and normalized using minmax normalization. f Violin plot of mRNA expression levels for genes upregulated?in tumor regions assigned with high aggressive scores (blue) versus those with?low scores (yellow). The top 10,000 spots from each group were shown. Boxes?within the violins represent the interquartile range (Q1-Q3) of the combined?groups, and circles inside the box represent median values. g, h Top enrichedbiological processes in tumor regions with (g) high and (h) low aggressiveness.Sizes of the circles represent the number of genes in each biological process, and?colors represent P values of enrichment. P values were determined using the?hypergeometric test and adjusted by the Benjamini-Hochberg procedure. Source?data are provided as a Source Data file.

圖 4 | 原位確定與預(yù)后相關(guān)的基因表達(dá)標(biāo)記。a 使用來自 TCGA 隊(duì)列的整張切片圖像訓(xùn)練了一個(gè)深度學(xué)習(xí)模型，以預(yù)測(cè)患者預(yù)后。用 H&E 染色的組織學(xué)圖像被裁剪成 56μm × 56μm 的片段。每個(gè)片段通過 ResNet-50 模塊轉(zhuǎn)換為特征向量（2048 × 1）。然后，這些特征向量通過一個(gè) Cox 回歸模塊映射到一個(gè)攻擊性評(píng)分上。每位患者的攻擊性評(píng)分平均用于驗(yàn)證。b 使用小組 (a) 中訓(xùn)練的圖像模型預(yù)測(cè)空間轉(zhuǎn)錄組學(xué)中斑點(diǎn)的攻擊性評(píng)分。c，d 在兩個(gè)腫瘤的空間轉(zhuǎn)錄組學(xué)隊(duì)列中可視化轉(zhuǎn)錄亞型和預(yù)測(cè)的攻擊性評(píng)分。攻擊性評(píng)分在每個(gè)樣本內(nèi)通過最小-最大規(guī)范化進(jìn)行規(guī)范化。e 邪惡斑點(diǎn)中位攻擊性評(píng)分的條形圖。所有腫瘤（n = 23）的攻擊性評(píng)分匯總并使用最小-最大規(guī)范化。f 小提琴圖顯示被分配高攻擊性評(píng)分（藍(lán)色）與低評(píng)分（黃色）的腫瘤區(qū)域中上調(diào)基因的 mRNA 表達(dá)水平。每組顯示前 10,000 個(gè)斑點(diǎn)。小提琴圖內(nèi)的盒子代表合并組的四分位數(shù)范圍（Q1-Q3），盒內(nèi)的圓圈代表中位數(shù)。g，h 在具有 (g) 高和 (h) 低攻擊性的腫瘤區(qū)域中富集的頂級(jí)生物過程。圓圈的大小表示每個(gè)生物過程中的基因數(shù)量，顏色表示富集的 P 值。P 值通過超幾何檢驗(yàn)確定，并通過 Benjamini-Hochberg 程序調(diào)整。來源數(shù)據(jù)以來源數(shù)據(jù)文件形式提供。

Fig. 5 | Screenshots of the GBM360 software. a Introductory page describing thefunctions of GBM360. The cell-type cartoons were created with BioRender.com.b Control panel for uploading histology images and configuring software settings.c Thumbnail of a histology image uploaded from the user. d Predictions and spatialvisualization of the cell-type distribution. The image was colored by transcriptionalsubtypes. e Predictions and visualization of regional aggressive scores. The imagewas colored by the aggressive score predicted at each patch. Red indicates highaggressiveness and blue indicates low aggressiveness. f–h Statistical analysis of thetranscriptional subtype distribution: (f) bar graph showing the transcriptionalsubtype fractions, (g) clustering coefficient for each subtype, and (h) twodimensional matrix showing the frequency of interactions between any two transcriptional subtypes.

圖 5 | GBM360軟件的屏幕截圖。a 入門頁面描述了GBM360的功能。細(xì)胞類型的卡通圖由BioRender.com創(chuàng)建。b 用于上傳組織學(xué)圖像和配置軟件設(shè)置的控制面板。c 用戶上傳的組織學(xué)圖像的縮略圖。d 預(yù)測(cè)和細(xì)胞類型分布的空間可視化。圖像按轉(zhuǎn)錄亞型進(jìn)行著色。e 預(yù)測(cè)和區(qū)域侵襲性得分的可視化。圖像按每個(gè)補(bǔ)丁預(yù)測(cè)的侵襲性得分進(jìn)行著色。紅色表示高侵襲性，藍(lán)色表示低侵襲性。f-h 轉(zhuǎn)錄亞型分布的統(tǒng)計(jì)分析：(f) 顯示轉(zhuǎn)錄亞型比例的條形圖，(g) 每個(gè)亞型的聚類系數(shù)，(h) 任意兩個(gè)轉(zhuǎn)錄亞型之間交互頻率的二維矩陣。

Table

表

Table 1 | Cox regression analysis showing the effect of sub type proportions on prognosis

表 1 | Cox 回歸分析顯示亞型比例對(duì)預(yù)后的影響

Table 2 | Cox regression analysis showing the effect of clus tering coefficient on prognosis

表 2 | Cox 回歸分析顯示聚類系數(shù)對(duì)預(yù)后的影響文章來源地址http://www.zghlxwxcb.cn/news/detail-860208.html

到了這里，關(guān)于文獻(xiàn)速遞：深度學(xué)習(xí)膠質(zhì)瘤診斷---空間細(xì)胞結(jié)構(gòu)預(yù)測(cè)膠質(zhì)母細(xì)胞瘤的預(yù)后的文章就介紹完了。如果您還想了解更多內(nèi)容，請(qǐng)?jiān)谟疑辖撬阉鱐OY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！