jenni-westoby/blueprint

Benchmarking pipeline published in https://genomebiology.biomedcentral.com/articles/10.1186/s13059-018-1571-5

Benchmarking pipeline for BLUEPRINT data

Prerequisites:

-virtualenv

-github account

-reference genome gtf and fasta files. The Mus musculus Ensembl release 89 genome was used in this study with ERCC sequences appended (see https://tools.thermofisher.com/content/sfs/manuals/cms_095048.txt).

-java version 1.8

-R version 3.4.4

To run the pipeline:

Execute ./wrapper.sh path/to/java path/to/ref/fasta path/to/ref/gtf

In practice it is unlikely that your machine will have the resources to run the entire pipeline in one go, so you will probably need to split up the wrapper script and run it in bits.

The pipeline automatically downloads the required data. In addition, a list of SRR accession codes can be found in SRR_Acc_List.txt.

As part of the pipeline, quality control steps are automatically carried out. For reference, these are the statistics used to filter the raw data:

Statistic	Name of statistic in table	Threshold
No. uniquely mapping reads	Unique	>8000000
No. of non-uniquely mapping reads	NonUnique	>350000
No. alignments	NumAlign	>8200000
No. of reads	NumReads	>4000000

These are the statistics used to filter the Polyester simulated data:

Statistic	Name of statistic in table	Threshold
No. of non-uniquely mapping reads	NonUnique	>250,000

In addition, the scater package was used to filter cells in which more than 10% of reads mapped to mitochondrial genes in both the raw and simulated data.