PO.BCS02.04 · 生物信息与计算
整合空间转录组学与MRI揭示与前列腺癌克隆异质性一致的独特放射-空间基因组图谱
Integrating spatial transcriptomics with MRI reveals distinct radio-spatial genomic profile concordant with prostate cancer clonal heterogeneity
作者与单位 Authors & Affiliations
摘要 Abstract
中文摘要
引言：前列腺癌（PCa）风险分层——即精确识别进展为转移性疾病风险最高的患者——仍然具有挑战性。整合临床数据、多参数MRI（mpMRI）影像组学和空间转录组学（ST）可能有助于识别与侵袭性PCa基因组特征相关的影像特征。
方法：对招募到一项试验（ISRCTN10046036）的患者的福尔马林固定石蜡包埋前列腺切除标本切片和转移性淋巴结进行了全器官空间转录组学分析（Visium™ v2，10x Genomics）。纳入了一例Gleason 4+4 PCa且有术前mpMRI的患者。这是一项子研究，是一个更大项目的一部分，该项目使用推断的克隆基因组分析（SpatialInferCNV）在十名男性中绘制前列腺癌克隆动态。图像处理涉及解剖分割以及与放射科医师和病理科医师的定性相关分析。使用ProsRegNet深度学习流程进行图像配准。使用SpatialStitcher（本地开发）将多个ST切片映射到一个共同坐标框架。使用PyRadiomics提取影像组学特征。
结果：代表超过85000个ST点（直径55μm）的八个ST切片被映射到T2轴位MRI。图像配准在前列腺包膜方面实现了0.861的DICE相似系数。尿道和BPH结节的标志点偏差分别为2.17mm和2.62mm。共选择了93个影像组学特征。一阶影像组学与ST点级组织学标注（根据良性 vs Gleason分级组（GG2、GG3和GG5））显著相关（p<0.001）。推断的克隆系统发育基因组分析揭示了轴位切片上的一个区域（X克隆）与淋巴结转移高度一致。该区域相比其他肿瘤和良性区域具有独特的z标准化纹理（特征：GLSZM、GLCM、GLDM、GLRLM、NGTDM）影像组学评分（p<0.001）。X克隆与其他肿瘤克隆之间在一阶影像组学（肉眼可见）方面未观察到差异（p = 0.21）。
结论：我们展示了前列腺内一种独特的纹理影像组学图谱，其与淋巴结转移具有系统发育相关性。该队列的进一步放射-空间基因组分析和验证分析正在进行中。这些可能为改进当前基于MRI的风险分层（用于随访和病灶靶向活检或局部治疗）提供见解。
查看英文原文 English abstract
Introduction: Prostate cancer (PCa) risk stratification, to precisely identify patients at greatest risk of progression to metastatic disease remains challenging. Integrating clinical data, multiparametric MRI (mpMRI) radiomics, and spatial transcriptomics (ST) may enable identification of imaging features linked to genomic signatures of aggressive PCa.
Methods: Organ-wide spatial transcriptomics (Visium TM v2, 10x Genomics) was performed on formalin-fixed paraffin-embedded prostatectomy sections and metastatic lymph nodes from patients recruited to a trial (ISRCTN10046036). A patient with Gleason 4+4 PCa and pre-operative mpMRI was included. This is a sub-study, part of a larger project mapping prostate cancer clonal dynamics in ten men using inferred clonal genomic analyses (SpatialInferCNV). Image processing involved anatomical segmentation and qualitative correlation with radiologist and pathologists. Image registration was performed using the ProsRegNet deep learning pipeline. Multiple ST sections were mapped to a common coordinate framework using SpatialStitcher (locally developed). Radiomic features were extracted using PyRadiomics.
Results: Eight ST sections representing over 85000 ST spots (55µm diameter) were mapped to T2-axial MRI. Image registration achieved a DICE similarity coefficient of 0.861 for prostate capsule. The landmark deviation for urethra and BPH nodule were 2.17mm and 2.62mm respectively. A total of 93 radiomic features were selected. First-order radiomics significantly correlated with ST spot-level histological annotation according benign vs Gleason grade group (GG2, GG3 and GG5) (p<0.001). Inferred clonal phylogenomic analysis revealed a region on the axial sections (X clone) with high concordance with lymph node metastasis. This region, harboured a distinct z-normalised textural (features: GLSZM, GLCM, GLDM, GLRLM, NGTDM) radiomic score compared to other tumour and benign regions (p<0.001). No difference in first order radiomics (visible to the eye) was observed between the X clone and other tumour clones (p = 0.21).
Conclusion: We show a distinct textural radiomic profile within the prostate that harbours a phylogenetic correlation with lymph node metastases. Further radio-spatial genomic profiling of the cohort and validation analyses are underway. These could offer insights into improving current MRI-based risk stratification for follow-up and lesion targeting for biopsy or focal therapy.
利益披露 Disclosure
T. Anbarasan, None..
S. Figiel, None..
M. Beesley, None..
W. Yin, None..
R. McPherson, None..
R. Colling, None..
F. Hamdy, None..
R. J. Bryant, None..
B. Papiez, None..
A. D. Lamb, None..
I. Mills, None.