The presence or absence alone of miRNA isoforms (isomiRs) successfully discriminate amongst the 32 TCGA cancer types

Abstract

Previously, we demonstrated that miRNA isoforms (isomiRs) are constitutive and their expression profiles depend on tissue, tissue state, and disease subtype. We have now extended our isomiR studies to The Cancer Genome Atlas (TCGA) repository. Specifically, we studied whether isomiR profiles can distinguish amongst the 32 cancers. We analyzed 10,271 datasets from 32 cancers and found 7,466 isomiRs from 807 miRNA hairpin-arms to be expressed above threshold.Using the top 20% most abundant isomiRs, we built a classifier that relied on "binary" isomiR profiles: isomiRs were simply represented as 'present' or 'absent' and, unlike previous methods, all knowledge about their expression levels was ignored. The classifier could label tumor samples with an average sensitivity of 93% and a False Discovery Rate of 3%. Notably, its ability to classify well persisted even when we reduced the set of used features (=isomiRs) by a factor of 10. A counterintuitive finding of our analysis is that the isomiRs and miRNA loci with the highest ability to classify tumors are not the ones that have been attracting the most research attention in the miRNA field. Our results provide a framework in which to study cancer-type-specific isomiRs and explore their potential uses as cancer biomarkers

Assessment of isomiR Discrimination Using Commercial qPCR Methods

“…In fact, a substantial background signal is observed in the 0|0 probe signal, both in synthetic and real cell contexts, before and after transfection of this isomiR ( Figure 2 , Figure 3 and Figure 4 ). If one were to attempt to probe −1|+1 levels using a qPCR probe, in order to confirm its absence from a sample and thereby heighten suspicion of a specific disease (see [ 16 ] for example), one would potentially encounter background signals that render the problem of confirming lack of a specific isomiR non-trivial. The reverse is true as well; in order to confirm the presence of a specific isomiR, one would need to be sure that any signal observed on qPCR captured the abundance of the isomiR itself, and not closely related sequences, which here are shown to produce non-trivial background noise.…”

“…We downloaded these data through the online data portal at: http://gdc-portal.nci.nih.gov. We then computed isomiR expression using the distributed miRNA isoform data, as in [11,16], and further subjected the computed reads to Threshold-seq, in order to determine a dynamic threshold for inclusion of isomiRs in downstream analysis [22]. Finally, we built expression matrices for isomiRs from each of the 32 TCGA projects, retaining only those isomiRs within a specific TCGA project that exceed Threshold-seq filtering level in at least one sample and retaining samples of any original type (specifically: tumor, normal, blood based, or metastasis).…”

“…Our previous analysis of 452 deep sequencing datasets from the GEUVADIS project [ 15 ] showed that isomiR profiles are able to classify individual samples based on characteristics, such as: sex, race, and population [ 11 ]. More recently, we showed that isomiRs can distinguish amongst tissue type and disease subtypes and that isoforms of the same miRNA can target largely non-overlapping groups of mRNAs [ 10 , 16 ].…”

MLW-gcForest: A Multi-Weighted gcForest Model for Cancer Subtype Classification by Methylation Data

“…The classification accuracy in Liao's study for BRCA and STAD is lower than that in our study, while the accuracy for LUAD is higher than that of our method. Telonis et al [41] evaluated the ability of isomiRs and used the top 20% abundant isomiRs to construct a binary classifier. The classifier could label tumor samples with 93% average sensitivity.…”

“…Methods Result AUC ACC Pre BRCA Liao [22] N/A 0.87 N/A Guo [20] N/A N/A 0.88 Telonis [41] N/A 0.91 N/A Li [42] 0.89 N/A N/A Sherafatian [43] N/A 0.89 0.90 MLW-gcForest 0.98 0.91 0.92 LUAD Liao [22] N/A 0.91 N/A Guo [20] N/A N/A 0.88 Telonis [41] N/A 0.86 N/A Podolsky [44] 0.92 N/A N/A Cai [19] N/A 0.85 0.86 MLW-gcForest 0.92 0.87 0.86 LIHC Guo [20] N/A N/A 0.82 Telonis [41] N/A 0.90 N/A Tan [45] 0.77 0.83 N/A Friemel [46] N/A 0.87 N/A MLW-gcForest 0.91 0.87 0.85 GBM Guo [20] N/A N/A 0.78 Lu [21] 0.92 0.88 N/A Ryu [47] 0.83 0.80 N/A MLW-gcForest 0.87 0.89 0.86 STAD Liao [22] N/A 0.84 N/A Telonis [41] N/A 0.85 N/A MLW-gcForest 0.88 0.87 0.87…”