Characterization of Continuous Transcriptional Heterogeneity in High-Risk Blastemal-Type Wilms’ Tumors Using Unsupervised Machine Learning

Abstract

Wilms’ tumors are pediatric malignancies that are thought to arise from faulty kidney development. They contain a wide range of poorly differentiated cell states resembling various distorted developmental stages of the fetal kidney, and as a result, differ between patients in a continuous manner that is not well understood. Here, we used three computational approaches to characterize this continuous heterogeneity in high-risk blastemal-type Wilms’ tumors. Using Pareto task inference, we show that the tumors form a triangle-shaped continuum in latent space that is bounded by three tumor archetypes with “stromal”, “blastemal”, and “epithelial” characteristics, which resemble the un-induced mesenchyme, the cap mesenchyme, and early epithelial structures of the fetal kidney. By fitting a generative probabilistic “grade of membership” model, we show that each tumor can be represented as a unique mixture of three hidden “topics” with blastemal, stromal, and epithelial characteristics. Likewise, cellular deconvolution allows us to represent each tumor in the continuum as a unique combination of fetal kidney-like cell states. These results highlight the relationship between Wilms’ tumors and kidney development, and we anticipate that they will pave the way for more quantitative strategies for tumor stratification and classification.

Characterization of alternative splicing in high-risk Wilms’ tumors

“…Likewise, we observed that tumors located near the stromal archetype have a high proportion of cells resembling renal fibroblasts. We performed similar analysis on a dataset of high-risk blastemal-type Wilms' tumors (treated according to SIOP protocols) published by Wegert et al [14] and found similar results [20] (see a detailed comparison in the supplementary information).…”

“…We then performed sequence alignment and obtained a gene expression matrix (Table S1). Using Principal Components Analysis (PCA) we found that high-risk Wilms' tumors form a continuum in gene expression latent space rather 6 than discrete well-separated clusters [19,20] (Figure 1A-F). To better understand this continuous heterogeneity, we used Pareto Task Inference [36,38] to calculate the vertices of the best fitting polytope encompassing all tumors and normal kidney samples in latent space.…”

“…In a previous study [19], we observed that Favorable Histology Wilms’ Tumors (treated according to the COG protocol) form a triangular shaped continuum in gene expression latent space, and that vertices of this triangle, which represent tumor “archetypes,” have blastemal, stromal, and epithelial characteristics, corresponding to the three main lineages of the developing fetal kidney. In a consecutive study [20] we found that this geometry is conserved also in high-risk tumors that were treated with chemotherapy prior to surgery (according to the SIOP protocol) but still contained a significant amount of remaining viable blastema [14], and used a probabilistic generative model [21] to represent each tumor as a mixture of three latent biological “topics” with stromal, epithelial, and blastemal characteristics.…”

Characterization of Alternative Splicing in High-Risk Wilms’ Tumors

“…In a previous study [20], we observed that favorable histology Wilms' tumors (treated according to the COG protocol) form a triangular shaped continuum in gene expression latent space, and that the vertices of this triangle, which represent tumor "archetypes," have blastemal, stromal, and epithelial characteristics, corresponding to the three main lineages of the developing fetal kidney. In a consecutive study [21], we found that this geometry is also conserved in high-risk tumors that were treated with chemotherapy prior to surgery (according to the SIOP protocol) but still contained a significant amount of remaining viable blastema [15], and used a probabilistic generative model [22] to represent each tumor as a mixture of three latent biological "topics" with stromal, epithelial, and blastemal characteristics.…”

“…We then performed sequence alignment and obtained a gene expression matrix (Table S1). Using Principal Component Analysis (PCA), we found that high-risk Wilms' tumors form a continuum in gene expression latent space rather than discrete well-separated clusters [20,21] (Figure 1A-F, Supplementary Figures S1-S6). To better understand this continuous heterogeneity, we used Pareto task inference [43,45] to calculate the vertices of the best fitting polytope encompassing all tumors and normal kidney samples in latent space.…”