Cluster analysis on high dimensional RNA-seq data with applications to cancer research - An evaluation study

Abstract

BackgroundClustering of gene expression data is widely used to identify novel subtypes of cancer. Plenty of clustering approaches have been proposed, but there is a lack of knowledge regarding their relative merits and how data characteristics influence the performance. We evaluate how cluster analysis choices affect the performance by studying four publicly available human cancer data sets: breast, brain, kidney and stomach cancer. In particular, we focus on how the sample size, distribution of subtypes and sample heterogeneity affect the performance.ResultsIn general, increasing the sample size had limited effect on the clustering performance, e.g. for the breast cancer data similar performance was obtained for n = 40 as for n = 330. The relative distribution of the subtypes had a noticeable effect on the ability to identify the disease subtypes and data with disproportionate cluster sizes turned out to be difficult to cluster. Both the choice of clustering method and selection method affected the ability to identify the subtypes, but the relative performance varied between data sets, making it difficult to rank the approaches. For some data sets, the performance was substantially higher when the clustering was based on data from only one sex compared to data from a mixed population. This suggests that homogeneous data are easier to cluster than heterogeneous data and that clustering males and females individually may be beneficial and increase the chance to detect novel subtypes. It was also observed that the performance often differed substantially between females and males.ConclusionsThe number of samples seems to have a limited effect on the performance while the heterogeneity, at least with respect to sex, is important for the performance. Hence, by analyzing the genders separately, the possible loss caused by having fewer samples could be outweighed by the benefit of a more homogeneous data.

Model-based feature selection and clustering of RNA-seq data for unsupervised subtype discovery

“…Thus, we generally observe a significant decrease in performance for larger K true , although FSCseq's performance is most robust to this effect. In addition, similarly sporadic performance of HC in RNA-seq was also observed previously (Vidman, Källberg and Rydén, 2019), suggesting unreliability of HC for clustering RNA-seq gene expression. Finally, although the MC methods performed similarly on average, the best performing method between these three transformations varied for each set of simulation conditions, reflecting the fact that the optimal transformation may not be known in advance.…”

Dynamic changes in fibrinogen and D-dimer levels in COVID-19 patients on nafamostat mesylate

“…Second, additional anticoagulants (i.e., intravenous heparin or direct oral anticoagulants), used on some patients, might have contributed to the dynamic changes in fibrinogen and D-dimer levels observed in this study. Third, while there is no definite rule for the minimum sample size for clustering approaches [29] and the indicative variables appear to be distributed appropriately, our small sample size requires extra caution when interpreting our findings. Further studies are needed to confirm our findings.…”

Persistent effects of pair bonding in lung cancer cell growth in monogamous Peromyscus californicus

“…Controls (C) cultured in the presence of fetal bovine serum (FBS) were also included. Unsupervised hierarchical clustering ( Vidman et al, 2019 ) indicated that the transcriptomes clustered well together according to the serum donors’ bonding history, except the virgin (V) group that exhibited the lowest discrimination ( Figure 3—figure supplement 2 ). Differential gene expression analysis was performed as described before by using the iDEP platform ( Ge et al, 2018 ).…”