Identifying New Potential Biomarkers in Adrenocortical Tumors Based on mRNA Expression Data Using Machine Learning

Abstract

Adrenocortical carcinoma (ACC) is a rare disease, associated with poor survival. Several “multiple-omics” studies characterizing ACC on a molecular level identified two different clusters correlating with patient survival (C1A and C1B). We here used the publicly available transcriptome data from the TCGA-ACC dataset (n = 79), applying machine learning (ML) methods to classify the ACC based on expression pattern in an unbiased manner. UMAP (uniform manifold approximation and projection)-based clustering resulted in two distinct groups, ACC-UMAP1 and ACC-UMAP2, that largely overlap with clusters C1B and C1A, respectively. However, subsequent use of random-forest-based learning revealed a set of new possible marker genes showing significant differential expression in the described clusters (e.g., SOAT1, EIF2A1). For validation purposes, we used a secondary dataset based on a previous study from our group, consisting of 4 normal adrenal glands and 52 benign and 7 malignant tumor samples. The results largely confirmed those obtained for the TCGA-ACC cohort. In addition, the ENSAT dataset showed a correlation between benign adrenocortical tumors and the good prognosis ACC cluster ACC-UMAP1/C1B. In conclusion, the use of ML approaches re-identified and redefined known prognostic ACC subgroups. On the other hand, the subsequent use of random-forest-based learning identified new possible prognostic marker genes for ACC.

Visual Clustering of Transcriptomic Data from Primary and Metastatic Tumors: Dependencies and Novel Pitfalls

“…These ndings were also in line with previous literature (Zhou et al, 2014). Additionally, using UMAP data dimension reduction with logarithmically transformed data of the TCGA-ACC (adrenocortical carcinoma) dataset revealed two clusters that closely matched the already known ACC subgroups (Marquardt et al, 2021a). This suggests that histopathological and cancer subgroupspeci c differences can be represented with a UMAP log10+1 approach, even though clusters seen within TCGA-KIPAN analysis were not completely subgroup-speci c, which can also be observed in t-SNE plot using unprocessed data.…”

“…Further details on the procedure are given elsewhere (Marquardt et al, 2021b). UMAP plots were generated based on an adapted UMAP approach as previously described (Marquardt et al, 2021a), using raw values as input. In brief, the squared pairwise Euclidean distance was used to calculate the distance between samples with a subsequent binary search for the optimal rho based on a xed number of 15 nearest neighbors.…”

Visual Clustering of Transcriptomic Data from Primary and Metastatic Tumors – Dependencies and Novel Pitfalls

“…Further details on the procedure are given elsewhere ( Marquardt et al, 2021b ). UMAP plots were generated based on an adapted UMAP approach as previously described ( Marquardt et al, 2021a ), using raw values as input. In brief, the squared pairwise Euclidean distance was used to calculate the distance between samples with a subsequent binary search for the optimal rho based on a fixed number of 15 nearest neighbors.…”

“…Visual clustering – based on data dimension reduction methods – is one potential approach to determine transcriptomic differences and proximities of different metastasis sites. The visualization methods mainly used for this purpose - t-SNE ( Laurens van der Maaten and Geoffrey Hinton, 2008 ) and UMAP ( McInnes et al, 09.02.2018 ) - have already been widely applied in the field of single cell sequencing ( Cillo et al, 2020; Puram et al, 2017; Zhang et al, 2021 ) but also bulk RNA-sequencing ( Marquardt et al, 2021a; Marquardt et al, 2021b; Zhao et al, 2020; Zheng et al, 2021 ) to visually separate transcriptionally similar cell populations from other diverging populations. Furthermore, there a recent studies also showing the critical impact of the initialization ( Kobak and Linderman, 2021 ) and used parameters ( Kobak et al, 2020 ) on dimension reduction methods.…”

New Insights in the Genetics and Genomics of Adrenocortical Tumors and Pheochromocytomas

“…Artificial-intelligence-based approaches can be used to uncover novel markers, as presented in two studies of this Special Issue. Marquardt et al revealed novel transcriptomic markers with prognostic relevance [ 6 ], whereas in the study by our research group (Turai et al) microRNA combination markers for adrenocortical malignancy with high diagnostic performance are presented [ 7 ]. A peculiar feature of adrenocortical cancer is represented by the high incidence of pediatric ACC in southern Brazil due to a founder mutation in the TP53 gene.…”