OpenAc4C: A gateway to decode the landscape, regulation and pathogenesis of N4-acetylcytidine (ac4C) epitranscriptome

OpenAc4C is the first comprehensive, freely accessible knowledgebase that integrates deep learning-identified ac4C sites and pathogenic variants across 33 species to facilitate the study of the ac4C epitranscriptome's landscape, regulation, and role in disease.

The Gist

Imagine your body's genetic code (DNA) as the master blueprint for building a human. But DNA doesn't do the actual construction work. Instead, it sends out copies of the instructions called RNA to the construction sites (cells).

For a long time, scientists thought these RNA copies were just plain, unedited photocopies of the blueprint. But recently, we discovered that these copies have little sticky notes or highlighters attached to them. These sticky notes change how the instructions are read, telling the cell to build things faster, slower, or to ignore certain parts. This field is called epitranscriptomics.

One specific type of sticky note is called ac4C. Think of it as a tiny "accelerator pedal" or a "stabilizer" attached to the RNA. When this pedal is pressed, it helps the cell make proteins more efficiently or keeps the RNA from falling apart too quickly.

The Problem: A Messy Library

The problem is that scientists have been finding these ac4C "pedals" all over the place, but the information is scattered.

• Some researchers found them in humans.
• Others found them in mice, plants, or even bacteria.
• Some found them using old, blurry cameras (low-resolution techniques), while others used high-definition microscopes (high-resolution techniques).
• There was no single place to look up where these pedals are, what they do, or how they relate to diseases like cancer.

It was like having a library where the books were scattered across different floors, in different languages, with no catalog.

The Solution: OpenAc4C

This paper introduces OpenAc4C, which is essentially a giant, super-smart digital library and map dedicated entirely to these ac4C sticky notes.

Here is what makes OpenAc4C special, explained simply:

1. The Ultimate Map (The "Google Maps" of RNA)
The researchers collected data from 33 different species (from humans to yeast to plants). They found over 536,000 ac4C sites.

• The Analogy: Imagine they built a global GPS system that shows you exactly where every single "accelerator pedal" is located on the RNA highways of 33 different countries.
• The Tech: They didn't just look at old data; they used Deep Learning (AI) to look at new, high-tech data from "Nanopore" sequencers. It's like using a satellite to find roads that ground-level maps missed.

2. The "What-If" Machine (Genetic Variants)
Sometimes, a typo in the DNA blueprint (a mutation) can accidentally delete an ac4C pedal or add a fake one.

• The Analogy: Imagine a construction worker accidentally gluing a "stop" sign where an "accelerator" should be. This causes the building to collapse or be built wrong.
• The Discovery: OpenAc4C scanned millions of genetic typos and found 536,000 of them that mess with these ac4C pedals. They found that about 4,700 of these typos are linked to serious human diseases, including cancer and hearing loss. It's like a detective connecting the dots between a broken instruction manual and a specific disease.

3. The User-Friendly Dashboard
The best part? You don't need to be a computer wizard to use it.

• The Analogy: Think of OpenAc4C as a video game interface for biology. You can type in a gene name, a disease, or a specific location, and it instantly shows you a colorful map, charts, and details.
• The Tools: It even has a "prediction engine." If you have a new RNA sequence and want to know if it has an ac4C pedal, you can upload it, and the AI will guess for you.

Why Does This Matter?

Think of ac4C as the volume knob on your radio.

• If the knob is turned up too high, the music (protein production) might be too loud and chaotic (cancer).
• If it's turned too low, the music is too quiet (disease).

Before this paper, scientists were trying to fix the volume by guessing where the knobs were. OpenAc4C gives them the exact manual with a map of every knob.

In short: This paper builds the first comprehensive "Google Maps" for a specific type of RNA modification. It helps scientists understand how tiny chemical changes control our health, how genetic errors cause disease, and provides a free, easy-to-use tool for researchers worldwide to solve these mysteries faster.

Technical Summary

1. Problem Statement

N4-acetylcytidine (ac4C) is a highly conserved RNA modification involved in gene expression, translation efficiency, and disease pathogenesis (e.g., cancer). Despite recent advancements in sequencing technologies (NGS and Nanopore) that have generated vast amounts of ac4C data, the field lacks a centralized, species-spanning resource. Existing databases (e.g., RMBase, RMDisease) cover multiple RNA modifications but provide only limited, fragmented information specifically regarding ac4C. Furthermore, there is no integrated platform that combines:

• Diverse sequencing data types (from low-resolution peaks to base-resolution sites).
• Third-generation sequencing (Oxford Nanopore) data.
• Systematic prediction of genetic variants affecting ac4C sites.
• Links between ac4C dysregulation and human disease phenotypes.

2. Methodology

The authors developed OpenAc4C, a comprehensive knowledgebase and web platform, through a multi-step data curation and computational pipeline:

• Data Collection & Integration:

  • NGS-based Data: Curated 226 raw sequencing samples from the NCBI GEO database covering four techniques: acRIP-seq (~150 bp resolution), PA-ac4C-seq, ac4C-seq, and RedaC:T-seq (base-resolution).
  • Standardization: Implemented a unified pipeline for acRIP-seq data (adapter trimming with Trim Galore, alignment with STAR, peak calling with exomePeak2) to ensure cross-study consistency.
  • ONT-based Data: Processed 42 FAST5 and 65 FASTQ files from 28 studies. Used the ELIGOS pipeline to identify modified cytosines based on base-calling errors, followed by a custom Deep Neural Network (DNN) model to specifically predict ac4C sites (scoring > 0.5, top 1% ranking). Additionally, utilized DirectRM for raw signal analysis.
  • Genetic Variants: Aggregated ~86 million germline and somatic variants from dbSNP, Ensembl, 1000 Genomes, and TCGA across seven species.

• Computational Prediction:

  • ac4C Site Prediction: Applied a deep learning model trained on experimentally validated NGS-ac4C sites to predict putative ac4C sites in ONT data and transcriptome-wide cytosines.
  • Variant Impact Analysis: Calculated an Association Level (AL) to quantify the impact of mutations on ac4C status. Variants were classified as ac4C-loss (removing a site) or ac4C-gain (creating a new site) by comparing predicted ac4C probabilities between wild-type and mutated sequences.
  • Disease Mapping: Cross-referenced ac4C-affecting variants with disease-associated TagSNPs from ClinVar, GWAS Catalog, and TCGA to identify pathogenic links.

• Platform Implementation:

  • Built using HTML, CSS, JavaScript, PHP, MySQL, and Bootstrap.
  • Integrated visualization tools (JBrowse, UCSC/Ensembl links) and analysis modules (R/Python scripts for backend processing).

3. Key Contributions

• First Dedicated ac4C Knowledgebase: OpenAc4C is the first resource exclusively designed to decode the ac4C epitranscriptome across diverse domains of life.
• Unprecedented Scale and Diversity:
  • Sites: Contains 536,745 ac4C sites across 33 species (including vertebrates, plants, fungi, bacteria, archaea, and viruses).
  • Techniques: Integrates four NGS-based profiling methods and, for the first time, a large collection of 88,936 putative ac4C sites derived from Oxford Nanopore direct RNA sequencing.
  • Variants: Catalogs 536,986 ac4C-affecting genetic variants across seven species.
• Disease Association Module: Identified 4,766 pathogenic ac4C-SNPs in humans linked to 1,280 disease phenotypes, including hereditary cancer syndromes and specific cancer types (e.g., Cholangiocarcinoma).
• Interactive Web Tools:
  • ac4C Finder: A deep learning-powered tool to predict putative ac4C sites from user-uploaded RNA sequences (FASTA) across 13 species.
  • SNP Analyzer: Allows users to upload VCF files to evaluate the impact of genetic variants on ac4C status (gain/loss prediction).

4. Key Results

• Data Reproducibility:
  • Validated the consistency of different sequencing methods. In humans, ~43.7% of RedaC:T-seq sites and ~60.7% of ac4C-seq sites overlapped with acRIP-seq peaks.
  • Permutation tests confirmed that the overlap between ONT-derived predictions and NGS data is significantly higher than random chance (e.g., 39.5% overlap in human), validating the utility of Nanopore data as a supplement.
• Disease Landscape:
  • Hereditary Diseases: 116 pathogenic variants (2.43%) were linked to hereditary cancer-predisposing syndromes; 57 to nonsyndromic hearing loss.
  • Cancer: Somatic analysis revealed 165,081 ac4C-associated variants across 33 TCGA cancer projects. Cholangiocarcinoma (TCGA-CHOL) showed the highest enrichment (8.40% of somatic mutations were ac4C-affecting), followed by Esophageal Carcinoma and Uterine Corpus Endometrial Carcinoma.
• Validation:
  • Independent wet-lab validation using LC-MS/MS confirmed ac4C presence in total RNA and Poly(A)+ RNA across four human cell lines.
  • RedaC:T-seq on SK-Hep-1 cells identified 6,304 sites, with 50% of newly identified base-resolution sites also found in acRIP-seq peaks, supporting the existence of mRNA ac4C.

5. Significance

OpenAc4C addresses a critical gap in epitranscriptomics by providing a unified, user-friendly gateway to the ac4C landscape. Its significance lies in:

1. Standardization: It harmonizes data from disparate sequencing technologies (NGS and Nanopore) and species, enabling comparative studies.
2. Mechanistic Insight: By linking genetic variants to ac4C dysregulation, it offers a new layer of understanding for disease pathogenesis, particularly in cancer and genetic disorders.
3. Resource Accessibility: The inclusion of deep learning tools allows researchers to in silico screen for ac4C sites and variant impacts before conducting expensive wet-lab experiments.
4. Future-Proofing: The platform is designed to accommodate emerging sequencing technologies and datasets, serving as a foundational resource for the rapidly evolving field of RNA acetylation research.

The database is freely accessible at www.rnamd.org/ac4cportal.