Subcellular SLAM seq: snakemake pipelines to fit subcellular T2C data

This repository contains three Snakemake pipelines to process metabolic labeling RNA sequencing data (fastq files) and fit a kinetic model to the processed data to estimate subcellular flow rates transcriptome-wide for each gene individually. This workflow was created while working on the paper Subcellular mRNA kinetic modeling reveals nuclear retention as rate-limiting by Steinbrecht et al. There are three main steps in the workflow:

• Transcript quantification (TPM normalized)
• T2C mutation counting and T2C data normalization
• Model fitting

Transcript quantification (TPM normalized)

Corresponding folder: tpm_quantification
This pipeline aligns read data to a genome and calculates transcript-per-million values using RSEM. TPM data from all samples are combined in a single file tpm_normalized_fraction_counts.csv. The mouse genome (GENCODE M14, same used in paper) is autmatically downloaded and used as reference geome. Configure the pipeline to your need in the config.yaml file.

T2C mutation counting and T2C data normalization

Corresponding folder: slam_seq_analysis
This pipeline contains a full-scale SLAM-seq analysis (in a GRAND-SLAM manner). First, sequencing data is aligned using STAR. Then, labeling-induced T2C mutations are counted using the custom-built CountT2C program, where T2C SNPs are excluded. Next, conversion rates are estimated based on the distribution of T2C mutation frequencies over the number of T counts in a read. Lastly, the ratio of new to total mRNA per intron and exon is estimated by normalizing the T2C counts with the conversion rates. The mouse genome (GENCODE M14, same used in paper) is autmatically downloaded and used as reference geome. Configure the pipeline to your need in the config.yaml file.

Model fitting

Corresponding folder: model_fit
This pipeline fits a kinetic model to the fully-processed subcellular labeling data using the python package lmfit. The are two models: 3-step and 4-step, each with its own Snakefile. Each Snakefile outputs two tables where each row corresponds to a specific gene. One table contains the best fit results for the kinetic parameters along with other fit statistics. The other table contains the varaiance of each kinetic parameter calculated from the ten best fit results.

Reproducibility

Instructions to run the the pipelines. The pipeline was tested on a SLURM-based high performance cluster.

• install conda (preferably miniforge)
• run conda env create -n snake -c bioconda snakemake=7.* to create a new conda environment with snakemake installed
• activate the environment with conda activate snake
• change the directory to the pipeline you want to run (e.g. cd slam_seq_analysis to do the SLAM-seq analysis)
• run the Snakemake pipleine with snakemake. The specific command to call snakemake can be found in the third line of each Snakefile.

Corresponding raw data

The fastq files for which this pipeline was made can be found on GEO under accession number GSE252199.

Notes

The folder brecht_profile contains a SLURM profile that can be used on SLURM-based HP clusters to execute the snakemake workflows.