I am interested in the potential for reproducible papers to be used the peer review process.

A highly-cited paper from 2010, "Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq", received 13 "contrasting" citations in scite.ai, indicating some groups found different

• mRNA splicing in METTL3 knockdown cells
• number of peaks
• peak location
• overall methylation levels

In order to study the possible role of bioinformatic approaches to this problem employed in Dominissini et al., I am proposing a "test of robustness", in which different tools, parameters, references, subsets are explored.

Rather than attempt to faithfully reproduce this paper using its original analysis stack, I have used a couple off-the-shelf pipelines to reanalyze this data:

• https://github.com/eQTL-Catalogue/rnaseq - The workflow processes raw data from FastQ inputs (FastQC, Trim Galore!), aligns the reads (STAR or HiSAT2), generates gene counts (featureCounts, StringTie) and performs extensive quality-control on the results (RSeQC, dupRadar, Preseq, edgeR, MultiQC).
• https://github.com/kingzhuky/meripseqpipe - MeRIP-seq analysis pipeline arranged multiple alignment tools, peakCalling tools, Merge Peaks' methods and methylation analysis methods.

To run:

adjust max_cpus and max_memory in meripseqpipe/conf/m6a.config as needed

gsutil -m cp -r gs://truwl-dominissini/SRP012098 .

gsutil -m cp -r gs://truwl-dominissini/refs .

gsutil -m cp -r gs://truwl-dominissini/SRP012098 .  (or s3://dominissini/SRP012098)
gsutil -m cp -r gs://truwl-dominissini/refs .       (or s3://dominissini/refs)



curl -s https://get.nextflow.io | bash
nextflow run meripseqpipe -profile stress,docker
nextflow run meripseqpipe -profile m6a,docker
nextflow run meripseqpipe -profile kd,docker

As a reproducible peer reviewer, you may...

• Manipulating parameters, data subsets
• Swapping out tools, reference data, or even primary data as needed
• Implementing further downstream analyses
• Implement something more similar to the original analysis
• Introduce new tools such as m6aviewer

...in order to evaluate the robustness of the results presented in this paper!

Data

m6a

Data from SRP012099 m6A RNA IP and Input for human RNA (untreated) - 7 files (4 IP/3 Input) metadata

Stress

Data from SRP012098 m6a RNA IP and Input for IFNg (200ng/ml) or HGF/SF (10 ng/ml) over night. Stress effects were tested in HepG2 cells by either 30 minutes incubation at 43ºC (heat shock) or UV irradiation) - 8 samples (IFN/HGF/HS/UV IP/Input) metadata

Knockdown

Data from SRP012096 METTL3_KD1 RNA-seq and mock controls - 5 files (2) - metadata

Mouse

Data from SRP012100

• SRR456934 GSM908344: Mouse_Liver_IP
• SRR456935 GSM908345: Mouse_Liver_Input

What has been done so far

• Code to generate manifests for meripseqpipe and rna-seq pipelines (see utils/metautils.py)
• A RNA-seq comparison of METTL3 knockdown & control HepG2 cells by Barry Digby
• Comparison of human METTL3 knockdown with mouse METTL3 KD in Xu et al by Sandeep Amberkar
• Meripseqpipe usage by Charlie Lonergan