TDMD

MiRNA cellullar levels are determined by the sum of two processes: biosynthesis and decay. It has recently been shown that targets with extended complementarity can induce miRNA degradation through a mechanism known as TDMD, Target-Directed miRNA Degradation. Compared to canonical miRNA:target pairing, TDMD is characterized by a different architecture that enables a target RNA to evade silencing and to destabilize its bound miRNA. Although natural TDMD-inducing targets are beginning to emerge, to date the number of predicted endogenous TDMD-targets and their possible impact on physiopathology in human cancers are yet to be defined.

What is TDMDfinder

TDMDfinder is an integrative online-platform for bioinformatically predicted High Confidence (HC) TDMD-interactions in mammalian genomes (Homo Sapiens and Mus Musculus).

My favorite organism is not included. What should I do?

TDMDfinder currently predicts TDMD-interactions in two organisms: human and mouse. If the organism you work with is not included, send us a request! If we receive a large number of enquiries in relation to the same organism, we will include its predicted TDMD-interactions in future versions of the platform.

How do I perform a search?

  1. Select your species of interest: Human or Mouse

  2. Submit a query. You can search TDMDfinder for predicted TDMD-pairs in any of the following three ways:

    1. enter the name of your miRNA of interest. In case of more than one match, all predicted pairs are returned;

    2. enter the Gene Symbol for your mRNA target of interest. In case of more than one match, all predicted pairs are returned;

    3. enter miRNA name and Gene Symbol for your TDMD-pair of interest. A single miRNA:target record will be presented directly.

  3. Set the “Additional filters” according to your preferences. By default, parameters are set on “Conserved” and “High Confidence” interactions. To refine the output, it is possible to apply, separately or in any combination, the following additional filters:

    1. conservation status of miRNA:target interaction, which allows to select conserved only (“Conserved”), non conserved only (“Non Conserved”), or both (“Both”), conserved and non conserved interactions together. Here, what is taken into account is the conservation of the miRNA site across vertebrate genomes, as described in Friedman et al. (2009);

    2. predicted (“Predicted”) and high confident (“High Confidence”) TDMD pairs, which allows to restrict the analysis only to those miRNA:target interactions that meet all the requirements for a potential or optimal TDMD effect, as defined by Simeone et al. (2021);

    3. experimentally supported interactions (“Validated interactions”), which allows to select miRNA:target pairs that have been annotated as experimentally validated direct interactions in the Human TarBse (Karagkouni et al., 2018) or identified via highthroughput approaches, such as HITS-CLIP (Helwak et al., 2013) and CLEAR-SEQ (Moore et al., 2015).

  4. Click on the “Search” button to perform your search.

Output

  1. The output is a table listing all identified interactions together with their most relevant features. The click on the “Column visibility” button allows the user to visualise additional information for all the resulting pairs shown in the output table. Table.1 gives more details on all the column headings that can be made visible in the results table.
Pair ID Unique pair identifier composed by miRBase ID, Gene Symbol ID, start and end positions on target 3’UTR.
Gene ID Click to link to Ensembl database.
Transcript ID Click to link to Ensembl database.
Gene Symbol Click to link to Ensembl database.
MiRBase ID Click to link to miRBase database.
MIMAT ID miRBase mature miRNA ID.
miR family miR family names according to TargetScan 7.1.
Seed match Seed type (8mer, 7mer-a1, and 7mer-m8).
Alignment Schematic representation of miRNA:target alignment. Target sequence on the top and miRNA sequence on the bottom.
UTR start Start position of predicted miRNA seed site in that species UTR.
UTR end End position of predicted miRNA seed site in that species UTR.
Conservation Conservation status of the predicted miRNA target site (Conserved = C, non conserved = nC).
Bulge Length of bulge.
3C consecutive Number of consecutive canonical Watson-Crick matches beyond seed region.
3C Consecutive G:U Number of consecutive matches (both, canonical Watson-Crick and not canonical wobble G:U base paring) beyond seed region.
3C type Nomenclature used to classify the 3C pairing region of each ‘Predicetd’ interaction. The classification taking into account the number of consecutive pairings and the contribution of G:U wobbles (3C_n7, 3C_n6+, 3C_n5+, 3C_<5+; a blank field is shown for the ‘not-Predicted’ couples). For more details see Fig.1 and Simeone et al., 2021.
Last four nt Number of matches involing the last four nucleotides of the microRNA sequence.
MFE The Minimum Free Energy (MFE) of RNA-RNA duplexes calculated by RNAhybrid tool.
MFE ratio Ratio between MFE and MFE value calculated for miRNA:perfect-TDMD-target, as described in Simeone et al. (2021).
MFE ratio category Nomenclature used to categorize the MFE ratio parameter of the all ‘Predicetd’ interactions (> 0.8/< 0.8).
Phylogenetic score Average of PhyloP basewise conservation scores calculated for 3C consecutive region.
Context score Targetscan 7.1 context score (Agarwal et al., 2015).
Weighted c. score Targetscan 7.1 weighted context score (Agarwal et al., 2015).
Predicetd TDMD miRNA:target pairs which match good requirements for TDMD (YES/NO), as described in Simeone et al. (2021).
High Confidence miRNA:target pairs which match the optimal requirements for TDMD (YES/NO), as described in Simeone et al. (2021).
Phylogenetic cons. If the TDMD-target MRE (microRNA degradation element) region is phylogenetically conserved (YES) or not (NO).
Validated If miRNA:target interactions are experimentally validated by cross-linking (CLASH), immunoprecipitation (CLEAR), or annotated as validated in DIANA-TarBaseV8 http://www.microrna.gr/tarbase (TARBASE). Otherwise not validated (NO).
Pancancer analysis Link to access to pancancer analysis session (only for Human). See below for more details.
Table.1: Explanation of columns in the results table.

Fig.1: Nomenclature used to classify the 3-Complementarity (3C) types.

  1. The result table can be easily downloaded in Comma Delimited (.csv) or Excel (.xlsx) format by clicking on “CSV” and “Excel”, respectively.

Pancancer analysis

The link in “Pancancer analysis” column provides a query entrance for a pancancer analysis session, which supplies for human interactions an integrative analysis across 21 cancer types (BLCA, BRCA, CESC, COAD, ESCA, HNSC, KIRC, KIRP, LGG, LIHC, LUAD, LUSC, OV, PAAD, PCPG, PRAD, READ, SARC, STAD, THCA, UCEC) retrieved from TCGA project.

Once you clicked on “run”, a result page with four major sections will appear for your favorite interaction.

  1. The first section is a table summarizing all the most relevant TDMD featurs for the selected pair.

  2. The second section, named “TCGA-Impact of TDMD on miRNA expression and Activity”, gives information for each cancer type about:

    • the total number of analyzed TCGA primary tumor samples with simultaneously miRNA-seq and RNA-seq data;

    • mean expression values of target and miRNA across cancer samples;

    • Spearman correlation coefficient computed between miRNA expression levels and TDMD target expression, relative R-squared (R^2) value, and statistics (p-value and FDR). If the interaction is an anticorrelated miRNA-target pair at a significant level ((p-value or/and FDR \(\leq\) 0.05), a star () is shown nearby the corresponding cancer type ID.

    • Spearman correlation coefficient computed between microRNA activity (estimated by using ActMir algorithm proposed by Lee et al. (2015)) and target gene expression values, and relative p-value;

    • a YES/NO column displaying “YES” if the two previously described correlation tests are in agreement (i.e. showing a statistically significant negative correlation);

    • graphical representations of target and miRNA expression profiles (boxplots on the top right and on the top left, respectively), correlation analysis (scatterplot on the bottom left), and ratio calculation (boxplot on the bottom right). For more details on ratio calculation see Marzi et al. (2016) and Ghini et al. (2018).

  3. The third section, “TCGA-Impact of CNV mRNA state on miRNA expression” (in abbreviated form, CNV Analysis), provides for the miRNA:target couples with paired target-CNV and miRNA-seq data an in-silico validation in order to test whether genetic deletions of the target show correlation with miRNA derepression. For the interaction of interest, a table summarizes across the 21 cancer types i) both the number of samples wild-type (WT) and harboring the deletion (DEL) of the target gene, ii) mean expression values of target and miRNA both in DEL group and in WT group, and iii) p-values from Student t-test and Wilcoxon test, is also provided. If in a given tumor type the genomic loss of TDMD-target correlates with higher miRNA expression level at a significat levels (p-value \(\leq\) 0.05 in at least one statistic test) and the expression conditions are fulfilled (i.e., mean expression of target is lower in DEL group of samples compared to WT pool of samples), a star () will appear nearby the corresponding cancer type ID. On the bottom of the table, a boxplot shows miRNA expression levels respect to the target CNV state, with relative statistics.

  4. The fourth section, “3UTR Analysis”, integrates 3P-seq (Nam et al., 2014) and TC3A data (Feng et al., 2017) to monitor loss of target:miRNA interactions due to 3’UTR shortening in cell lines and cancer datasets. Information is provided in a tabular and graphical format when data are available. Table.2 gives more information about columns in the 3UTR Analysis results table.

AIR 3UTR
Lenght Total lenght of the 3’UTR as defined in TargetScan 7.1.
Start-End Position of the miRNA seed within 3’UTR.
Range (0-100) Affected isoform ratio. The ratio indicates for each miRNA target site the fraction of mRNA transcripts containing the site as measured by 3P-seq in Nam et al. (2014).
Tags Expressed tags of the most abundant isoform for the transcript.
Class According to the AIR range, four classes are defined: 100, very good; 75-100, good; 50-75, mid; 25-50, poor; 0-25, very poor.
TCGA Cancer datasets
N Datasets Number of TCGA datasets for which we could retrieve data from.
N Samples Number of samples (pancancer) with PDUIs information.
Sample with MRE loss Number of samples (pancancer) in which the PDUI value is below 0.5 (meaning that more than 50% of transcripts are in the short form AND the short form does not contain the miRNA binding site.
Fraction in pantumor N Sample with MRE loss / N samples pantumor. Represent the pancancer frequency of MRE loss.
Fraction in tumor by tumor Average value of N Sample with MRE loss / N samples calculated in each individual tumor.
PDUI average Average value of PDUI across all samples.
TC3A Class According to the pantumor frequency of MRE loss, three classes are defined: rare, <10% of cases; frequent, between 10% and 50% of cases; very frequent, >50% of cases.
Table.2: Explanation of columns in the 3UTR Analysis results table.

Examples Here, we take as an example “hsa-miR-30c-5p:SERPINE1” pair, which has been reported to be involved in a TDMD-interaction (Ghini et al., 2018).

How many TDMD-targets could interact with hsa-miR-30c-5p in Human genome in Conserved High Confidence set?

  1. Select the right “Species”: Human (Fig.2);
  2. Enter hsa-miR-30c-5p (official miRBase ID) in the miRNA ID box (Fig.2);
  3. Leave the “Additional filters” on default settings, i.e. Conserved & High Confidence (Fig.2);
  4. Click on the Search button (Fig.2);
drawing

Fig.2: hsa-miR-30c-5p target interactions search.

  1. Browse the table results.

What is the relationship between hsa-miR-30c-5p and SERPINE1 in Breast Invasive Carcinoma (BRCA)?

  1. Select the “Species”: Human (Fig.3);
  2. Put hsa-miR-30c-5p and SERPINE1 in the opened boxes (Fig.3);
  3. Select the “Additional filters” of interest. Here we use Both (Fig.3);
  4. Click on the Search button (Fig.3);
  5. Click on the run link in “Pancancer analysis” column (Fig.3);
drawing

Fig.3: hsa-miR-30c-5p:SERPINE1 interactions search.

  1. Select the cancer type BRCA in the “TCGA Expression Analysis” table or from the underlying drop-down menu (Fig.4)
drawing

Fig.4: Query of hsa-miR-30c-5p:SERPINE1 in BRCA.

  1. Browse the results. Click a point on each graph to view more details about the samples and/or the expression distributions (Fig.5).
drawing

Fig.5: Result page of pancancer expression analysis (focus on BRCA) of hsa-miR-30c-5p:SERPINE1 interaction.


hsa-miR-30c-5p:SERPINE1 is a High Confidence pair. What is the genomic impact of CNV status of the gene target on miRNA expression in BRCA?

  1. Select the “Species”: Human (Fig.3);
  2. Put hsa-miR-30c-5p and SERPINE1 in the opened boxes (Fig.3);
  3. Select the “Additional filters” of interest. Here we use Both (Fig.3);
  4. Click on the Search button (Fig.3);
  5. Click on the run link in “Pancancer analysis” column (Fig.3);
  6. Select the cancer type BRCA in the “TCGA Expression Analysis” table or from the underlying drop-down menu (Fig.4);
  7. Browse more details in “TCGA Impact of CNV mRNA state on miRNA expression” section. If you click a point on the plot you can view more expression details about the two analyzed patient cohorts (DEL vs WT) (Fig.6).
drawing

Fig.6: Result of CNV analysis for the couple hsa-miR-30c-5p:SERPINE1. On the top the impact of SERPINE1 CN state on hsa-miR-30c-5p expression level across all cancer types. On the bottom a focus in BRCA tumor type for the couple of interest.


Cancer cells can express transcripts with systematically shorter 3′ UTRs compared to normal cells. What about the 3′ UTR‐based regulation for hsa-miR-30c-5p:SERPINE1 interaction in cancer disease?

  1. Select the “Species”: Human (Fig.3);
  2. Put hsa-miR-30c-5p and SERPINE1 in the opened boxes (Fig.3);
  3. Select the “Additional filters” of interest. Here we use Both (Fig.3);
  4. Click on the Search button (Fig.3);
  5. Click on the run link in “Pancancer analysis” column (Fig.3);
  6. Scroll down the webpage or use the related jump link (3UTR Analysis) (Fig.7);
drawing

Fig.7: Focus on “3UTR Analysis” jump link.

  1. When available, browse results in “3UTR Analysis” section (Fig.8). drawing

    Fig.8: Focus on results of the “3UTR Analysis” for hsa-miR-30c-5p:SERPINE1 interaction. jump link.