Isoform Workflow

Requirements

• GNU Make
• Bash Shell
• jq
• curl
• R (packages: jsonlite)
• CESAR 2.0 (https://github.com/hillerlab/CESAR2.0)

Installation

This workflow can be downloaded from GitHub using git clone.

git clone https://github.com/CartwrightLab/isoform_workflow.git

Usage

This workflow is built around GNU Make.

cd isoform_workflow
make data/fasta_cds/AREG.fasta

This command downloads information from Ensembl for a human gene. It will also download the genomic sequences of the human gene and all the one-to-one orthologs identified in other primates. It then constructs an alignment of the human canonical CDS with the other species.

Directory structure

• fasta_cds: Multiple sequence alignments of CDS sequences.
• fasta_raw: Raw genomic sequences downloaded from Ensembl.
• gene_info: Information about canonical isoforms.
• gene_info_raw: Raw gene information downloaded from Ensembl.
• homology: Information about one-to-one orthologous genes.
• homology_raw: Raw homology information downloaded from Ensembl.

Potential Issues and Improvements

The CESAR alignments are currently used as is and low quality "hits" are not filtered out.

CESAR can take a while to run, but provides better results on our problem than lastz, nucmer, and minimap2. It's possible that exonerate might be able to replace CESAR, or that we could rethink our approach to using lastz.

The results in fasta_cds are quick, reference-based alignments. A full tree-based, codon-aware multiple sequence aligner would produce more optimal results. But would likely require masking of "frameshifts" relative to the reference first. Without writing a new aligner, we can probably do fasta_cds -> coati -> masking -> prank. However, coati might be redundant because CESAR is already codon-aware.