Notes about datasets

1. Atlas_2022_tissue: Next-generation sequencing set from miRNATissueAtlas2. Contains only physiological tissues from Human and Mouse.

2. Atlas_2016_tissue: microarray data from original miRNATisuseAtlas. Contans only physiological tissues from Human.

3. Atlas_2025_tissue: Next-generation sequencing set combining miRNATissueAtlas2 and data from physiological tissue samples in miRNATissueAtlas 2025.

4. Atlas_2025_cell_line: Next-generation sequencing set from only cell lines in miRNATissueAtlas 2025.

5. Atlas_2025_EV: Next-generation sequencing set from only EV in miRNATissueAtlas 2025.

Specificity

The specificity page summarizes the results for the tissue specificity index calculation. (See mirnatissueatlas2: an update to the human miRNA tissue atlas). There are 4 primary sets for which we compute TSI values.

The entire table is available for download as a csv file using the "Download TSI summary" button.

Columns:

• acc: Search non-coding RNA.
• species: filter for species, options: 'hsa' for H. sapiens and 'mmu' for M. musculus
• type: type of non-coding RNA, options: 'mirna', 'snrna', 'lincrna', 'snorna', 'pirna', 'scarna', 'trna', 'rrna', 'miscrna'
• [dataset] expressed samples count: number of samples in the dataset where the non-coding RNA was expressed. Using the min and max filters, all non-coding RNA that were expressed in that range of samples in that dataset can be filtered.
• [dataset] TSI: tissue specificity index calculated using the formula mentioned in the mirnatissueatlas2 manuscript, taking 'organ' as category by which to group the samples. Using the min and max filters, all non-coding RNA that had a TSI in that range in that dataset can be filtered.

As there are over 100k non-coding RNA, the table is split into multiple pages which can be browsed using the pagination at the bottom of the table.

ncRNA patterns

The expression patterns in different organs for each non-coding RNA can be explored using the ncRNA patterns page. Here there is the option to select 'species', 'source', 'norm', 'ncrna' and 'acc' which corresponds to the accession for the non-coding RNA. All data sources or species do not have the same combination of normalization method or ncRNA available. Not all accessions correspond to all classes of non-coding RNA.

Therefore, the form is designed to be self-filtering: only the set of options possible for each field is shown. For example, there is no mouse Atlas_2016_tissue dataset, therefore, picking 'mmu' in the species section will not allow 'Atlas_2016_tissue' to be selected as an option.

Clicking the submit button will generate box plots of the expression of that non-coding RNA in each of the organs in the selected subset. The card at the top will contain The unique name, common name and external links to miRBase, RNA central and if present, miRGeneDB.

A table with the isomir information is also generated with links that will lead to a boxplot of the isomir expression in different organs.

Tissues

The tissues form allows querying of the average and standard deviation of the non-coding RNA subset for corresponding "species", "source", "norm" in specific "tissue". Similar to ncRNA patterns, the form is self-filtering. Conveniently, this allows the search of tissues corresponding to organ_system. Note, currently we group and subgroup the data according to "organ_system" and "tissue". It can be argued that certain "tissue" can belong to multiple "organ_system" and that certain "organ_system" can themselves be classified into a larger system. We are currently working to create such a hierarchical mapping.

Selecting the desired field will result in a table containing the accession, average value and standard deviation for all non-coding RNA in that particular tissue. Since there can be thousands of non-coding RNA, the pages at the bottom can be used to browse through the data.

Alternatively, the tables can be downloadable as parquet files.

miRNA Correlations

The miRNA correlations tab allows selection of 'Species', 'Source', 'rpmm' allows the identification of the top 10 most correlated within a subset using Spearman Rank Correlation. Scatterplots of the top 10 most correlated miRNA are also generated after submitting the data.

Note: Currently only miRNA correlation is enabled due to the massive volume of data available.

Network Analysis

Network analysis has a form where the 'species', 'source', 'norm', 'ncrna' can be selected. Following this multiple options from 'Acc', 'Tissue' and 'Organ system' can be selected. Using the selected accessions, tissue and organ system, we create a graph based on whether or not the tissue is connected to the organ system and whether the non-coding RNA is connected to the tissue. We determine a connection based on the options 'Any' and 'All' and a user defined threshold number.

The threshold serves as the cutoff point to determine if an edge will be formed between a non-coding RNA and a tissue: if the average value of a non-coding RNA is greater or equal to the threshold, it will form form an edge, otherwise not. 'Any' and 'all' are used to filter the non-coding RNA based on this binary value.

'Any' selects any non-coding RNA that is expressed in each of the tissues that have been selected. Therefore, it is more restrictive. 'All' selects a non-coding RNA that is expressed in at least one of the tissues that have been selected. Therefore, it is more permissive.

Custom Heatmap

Custom heatmap allows the selection of 'species', 'source', 'norm' and 'ncRNA' along with multiple values from 'Accession' and 'Tissue' which can be further filtered by the 'Organ system'. Clicking the submit button results in the creation of a heatmap with the samples corresponding to each Tissue on the column and each non-coding RNA on the row using clustergrammer2.

Note, to keep the visualization neat, the number of non-coding RNA accession is limited to 100 and the number of tissues is limited to 10. Nevertheless, there might be a large number of samples in the selected tissues, making the visualization of columns difficult.

miEAA Analysis

miEAA analysis allows the selection of miRNA and calculation of Over-representation analysis using miRNA Enrichment Analysis and Annotation (miEAA). Clicking the submit buttons uses the miEAA API to create a job and takes the user off_site.

Downloads

The tables used to visualize the data for both legacy version 2 and updated miRNATissueAtlas 2025 is available as matrices. miRNATissueAtlas 2 tables are available as csv files. However, due to the massive size of the data for miRNATissueAtlas 2025, currently the datasets are available only as h5ad files, which is a highly compressed file format for large datasets.