TEExplorer: A Web Portal to Investigate TE-Epigenome Associations Across Human Cell Types

The authors developed TEExplorer, an interactive web portal that provides accessible visualization of transposable element and histone mark associations across 57 human cell types from the EpiATLAS dataset while enabling users to upload and compare their own ChIP-seq data.

The Gist

Imagine your genome (your body's instruction manual) is a massive, 3-billion-page library. For a long time, scientists thought about half of this library was just "junk" or "noise"—pages that were copied, pasted, and scattered everywhere by ancient viruses and genetic parasites called Transposable Elements (TEs).

However, recent research suggests this "junk" isn't actually trash. It's more like a hidden layer of sticky notes, highlighters, and post-it reminders that tell the library how to function. These TEs can act as switches that turn genes on or off, influencing everything from how your immune system fights a virus to how your brain develops.

The problem? There are so many of these "sticky notes" (over 6 million measurements in one dataset alone!) that it's impossible for a human to read them all and figure out which ones matter. It's like trying to find a specific sentence in a library where every book is written in a different language and half the pages are missing.

Enter TEExplorer: The "Google Maps" for Genetic Sticky Notes.

This paper introduces a new web tool called TEExplorer. Think of it as a user-friendly dashboard that helps scientists navigate this chaotic genetic library. Here is how it works, broken down into simple concepts:

1. The Big Picture (The "Overview" Map)

Imagine you want to know which neighborhoods in a city have the most coffee shops. TEExplorer lets you zoom out and see a map of the entire genome. It shows you:

• Which "families" of TEs (like the "Alu" family or the "L1" family) are hanging out near specific genetic switches (called histone marks).
• Which cell types (like brain cells vs. blood cells) use these switches differently.

The Analogy: It's like a weather map. Instead of showing rain and sun, it shows where specific genetic "storms" (TEs) are gathering around specific "cities" (cell types) to see if they are causing a "flood" (gene activation) or a "drought" (gene silencing).

2. The Detective Work (The "Subfamily" View)

Sometimes, you need to look closer. Maybe you know the "Alu" family is important, but you want to know which specific members of that family are doing the work.

• TEExplorer lets you drill down. You can pick a specific TE family and a specific cell type.
• It then creates a heat map (a colorful grid) showing exactly which sub-families are "enriched" (showing up more than expected by chance) and which are "depleted" (showing up less).

The Analogy: If the "Alu" family is a large sports team, this feature lets you look at the roster to see if it's the "AluY" players or the "AluSz" players who are actually scoring the goals in the brain cells.

3. Bring Your Own Data (The "Upload" Feature)

This is the coolest part. Scientists often have their own experiments (like studying how flu viruses affect immune cells). Before, they had to do complex math to compare their results to the massive global database.

• With TEExplorer, a scientist can simply upload their own file (like a list of genetic coordinates).
• The tool instantly compares their data against the massive database of 4,600+ existing samples.
• It tells them: "Hey, your flu-infected cells look just like the standard immune cells, except for this one specific genetic element that is acting strangely."

The Analogy: It's like taking a photo of your own house and uploading it to a real estate app. The app instantly compares your house to thousands of others in the neighborhood and says, "Your house is very similar to the others, but you have a unique blue door that no one else has."

Why Does This Matter?

The authors tested this tool by looking at cells infected with the Flu virus. They wanted to see if the virus changed how the "sticky notes" (TEs) were arranged.

• The tool quickly showed that while most things looked normal, a specific type of genetic element (called THE1B) was acting differently in the infected cells compared to healthy ones.
• This confirmed previous findings but did it in a way that didn't require a PhD in computer science to run the analysis.

The Bottom Line

TEExplorer is a bridge. It takes a massive, intimidating mountain of genetic data and turns it into a simple, interactive playground. It allows researchers who aren't experts in "jumping genes" to easily ask: "What are these genetic parasites doing in my cells?" and get an answer in seconds.

It's not just a database; it's a flashlight that helps us see the hidden instructions in our DNA that might hold the keys to understanding disease, immunity, and human development.

Technical Summary

1. Problem Statement

Transposable elements (TEs) constitute approximately 50% of the human genome and play critical roles in gene regulation, development, immunity, and disease. However, analyzing TEs is challenging due to their repetitive nature and the complexity of integrating them with epigenomic data.

• Data Volume: Previous comprehensive analyses of the International Human Epigenome Consortium (IHEC) EpiATLAS dataset (4,614 ChIP-seq samples across 57 cell types and 6 histone marks) generated over 6.5 million measurements of TE/histone mark/cell type associations.
• Accessibility Gap: While the raw data and initial analysis existed, the volume of results made it difficult for researchers to navigate specific associations or integrate their own data with the existing atlas. Existing tools (e.g., UCSC Repeat Browser, WashU Repeat Browser) lacked the breadth of cell types, histone marks, or the ability to perform comparative enrichment analysis against a large, uniformly processed dataset.

2. Methodology

The authors developed TEExplorer, a web portal designed to visualize and analyze TE-epigenome associations.

Data Processing & Backend

• Source Data: Utilized the IHEC EpiATLAS dataset containing 4,614 ChIP-seq samples covering 6 histone marks (H3K27ac, H3K4me3, H3K4me1, H3K36me3, H3K27me3, H3K9me3) across 57 human cell types.
• TE Annotation: Peaks were overlapped with TE coordinates from RepeatMasker (hg38).
• Enrichment Calculation:
  • Observed Overlap: Counted peaks overlapping TEs.
  • Expected Overlap (Random Control): Generated 1,000 simulations of random 200bp regions per sample, maintaining the same distance-to-TSS distribution as the real peaks.
  • Metrics: Calculated Observed – Expected counts, Fold Change (Observed/Expected), and statistical significance (p < 0.005 based on 995/1000 simulations).
• Database Structure: Data was pre-aggregated into two SQLite databases:
  1. Subfamily Level: Granular data for 1,426 TE subfamilies.
  2. Family Level: Aggregated data for 60 TE families.
• Quality Control (UMAP): To compare user data with the atlas, the authors created a dimensionality reduction model. They binned the genome into 10kb windows, selected the 20,000 most variable windows, and trained a UMAP model on the EpiATLAS samples.

Software Architecture

• Stack: Built using R, Shiny, and ShinyDashboard.
• Visualization: Uses ggplot2, plotly (interactive), and ComplexHeatmap.
• Performance: SQLite databases allow for rapid querying without live computation of the massive underlying datasets.

3. Key Contributions & Features

TEExplorer offers three primary interactive sections:

1. TE Overview:

   • Provides a broad view of TE family enrichment across all cell types and histone marks.
   • Features interactive bar plots for overlap percentages and heatmaps for subfamily enrichment (Observed – Expected).
   • Allows filtering by specific TE families, assays, or cell types.

2. TE Subfamilies:

   • Enables deep-dive analysis into specific TE families (e.g., Alu, L1).
   • Displays boxplots and heatmaps of enrichment for specific subfamilies across cell types.
   • Includes an option to view "depleted" TEs (negative enrichment).

3. Import (User Data Analysis):

   • Allows users to upload their own ChIP-seq BED files (up to 50MB).
   • Automated Processing: Resizes peaks to 200bp (to match EpiATLAS), calculates TE overlaps, and projects the sample onto the pre-trained UMAP space to verify cell type/assay clustering.
   • Comparative Analysis: Compares user data against the EpiATLAS mean and random controls using three metrics: Observed–Expected, Observed Count, and Fold Change.
   • Highlighting: Identifies extreme outliers (most enriched/depleted) and subfamilies with the largest deviation from the EpiATLAS mean.

4. Results & Validation

The authors validated the tool using two distinct use cases:

• Validation with EpiATLAS Data:

  • Re-uploading a T-cell H3K9me3 sample from the EpiATLAS dataset confirmed that the tool correctly clustered the sample with its biological replicates in the UMAP plot.
  • The tool accurately reproduced known enrichment patterns (e.g., L1 elements enriched in H3K9me3; Alu elements enriched in H3K27ac).

• Case Study: Influenza A Infected Macrophages:

  • The authors analyzed an independent dataset of monocyte-derived macrophages (MDMs) infected with Influenza A (FLU) vs. non-infected (NI).
  • Findings:
    • Global TE overlap was consistent with EpiATLAS macrophage data (~34%).
    • Differential Enrichment: The tool detected specific subfamily differences. Notably, ALR/Alpha repeats were more enriched in the uploaded samples than in EpiATLAS, with further enrichment in FLU samples compared to NI.
    • THE1B Subfamily: The tool identified higher enrichment of THE1B (within the ERVL-MaLR family) in FLU samples compared to NI, recovering a result previously published by Chen et al.
  • This demonstrated the tool's ability to detect biologically relevant, subtle differences in TE regulation during infection without requiring custom scripting.

5. Significance

• Democratization of TE Analysis: TEExplorer lowers the barrier to entry for researchers without specialized TE expertise, allowing them to explore complex TE-epigenome associations via a standard web browser.
• Hypothesis Generation: By visualizing over 6.5 million pre-calculated associations, the tool helps researchers identify TEs that may have been overlooked as causal elements in their experiments.
• Benchmarking: It provides a standardized framework for comparing new ChIP-seq datasets against a massive, uniformly processed reference (EpiATLAS), facilitating the identification of cell-type-specific or condition-specific TE regulatory events.
• Limitations: The tool relies on reference genome annotations (missing polymorphic TEs) and excludes multi-mapping reads (potentially underestimating younger TEs). However, it provides mappability estimates to help users interpret these constraints.

Availability: The portal is publicly accessible at https://teexplorer.c3g.sd4h.ca.