Enhancing Non-Coding DNA Alignment Using Gene Expression Data

Introduction

The alignment of non-coding DNA sequences presents significant challenges due to their variability and lack of conservation compared to protein-coding regions. This project explores an innovative approach to enhancing the alignment of non-coding DNA, particularly promoter regions, by incorporating gene expression data. The ultimate goal is to develop a more informative similarity measure that better reflects evolutionary relationships and functional similarities.

Motivation

Alignment methods traditionally rely on sequence similarity to determine evolutionary and functional relationships. However, for non-coding DNA sequences such as promoters, conventional alignment techniques often fail to provide meaningful insights. As an example, let us consider orthologs (= same gene in different species) . Very closely related, shared origin and presumably same function.

• Current alignment methods show low accuracy, as observed in a baseline comparison:
  • Protein-coding orthologs: Standard alignment methods correctly identify orthologous relationships with an accuracy of 99.7%.
  • Promoter sequences: The same alignment methods achieve only 43.5% accuracy in identifying true orthologous relationships.

By incorporating gene expression data, this project aims to achieve a better representation of the promoter sequences create a similarity measure that improves alignment accuracy, making it easier to distinguish between related and unrelated promoter sequences.

3. Gene Expression Similarity:
   • Gene expression profiles will be compared across tissues in human and mouse to assess functional relatedness.
   • The hypothesis is that promoters regulating genes with similar expression patterns are more likely to be functionally related.

Dataset

The starting dataset includes human and mouse one-to-one orthologs within the kinase family. The key components of the dataset are:

• Mapping Information: Ortholog relationships between human and mouse genes.
• Raw Promoter Sequences: Extracted upstream sequences of orthologous genes.
• Protein Domain Sequences: To assess the similarity of downstream coding regions.
• Gene Expression Data: Expression profiles across various tissues in both species.

Repository Structure

The repository follows an organized structure to facilitate reproducibility and scalability:

/promoter_expression
│-- data/                  # Raw and processed data (sequences, mapping info, expression data)
|   |-- db_ids/            # Mapped ensemble gene ids (mouse/human orthologs)
|   |-- expression/        # Gene counts
|   |-- sequence/          # Promoter and protein sequences (human/mouse)
│-- modules/               # Nextflow modules for core tasks
│   │-- pairwise_alignment/  # Align promoter or protein sequences
│   │-- transform_expression/ # Transform expression data using several methods 
│   │-- exp_similarity/ # Compute expression-based similarity (cosine and pearson)
|   |-- combine_similarity/ # Combine sequence and expression similarity
|   |-- label_pairs/ # Label positive/negative pairs for training using different criteria
│-- workflows/             # Workflow scripts integrating multiple modules
│-- scripts/               # Additional scripts (data preprocessing, sequence retrieval, etc.)
│-- exploratory_analysis/  # Exploratory data analysis scripts & plots
│-- results/               # Output data (final similarity matrices, figures, reports)
│-- README.rd              # Project documentation

Implementation Plan

1. Development of Nextflow Modules

To ensure modularity and reproducibility, core functionalities will be implemented as Nextflow modules:

• Promoter Alignment Module: Aligns promoter sequences using traditional sequence-based methods.
• Protein Domain Alignment Module: Aligns protein domains to provide baseline evolutionary relationships.
• Expression Similarity Module: Computes gene expression similarity across tissues and species.

2. Workflow Integration

A set of workflows will be developed to automate the full analysis pipeline, allowing for flexibility in selecting different datasets and similarity metrics. These workflows will:

• Process raw data and prepare inputs for alignment.
• Perform multiple alignment strategies and evaluate their performance.
• Generate final similarity matrices integrating sequence-based and expression-based measures.

3. Exploratory Analysis

To assess the effectiveness of the proposed approach, exploratory analysis will be conducted. This includes:

• Visualization of similarity distributions using various metrics.
• Comparative analysis between sequence-based and expression-based similarity.
• Evaluation of alignment informativeness in distinguishing true orthologs from unrelated sequences.

Expected Challenges

• Developing scalable Nextflow modules to handle large genomic datasets efficiently.
• Defining an optimal similarity metric for expression-based alignment.
• Ensuring generalizability beyond the kinase family dataset to other gene families.

Next Steps

1. Implement and test Nextflow modules for promoter alignment, protein domain alignment, and expression similarity.
2. Optimize similarity metrics to improve classification accuracy of promoter alignments.
3. Perform comparative analyses between sequence-based and expression-based methods.
4. Expand the dataset to include additional gene families and species.

Contributions & Collaboration

This project is a work in progress, and contributions from others in the field are highly welcome. If you have expertise in:

• Genomic sequence alignment
• Gene expression analysis
• Computational pipeline development

please feel free to contribute! If you have suggestions or would like to collaborate, contact [Your Name] at [Your Email].

---

Author: Cristina Araiz Affiliation: Center for Genomic Regulation Contact: cristina.araiz@crg.eu