RapMap: reference transcriptome for simulated reads

Question

I am very new to bioinformatics and trying to repeat the benchmark in the RapMap paper with an experimental tool working in a similar but different fashion.

In the paper (taken from their github) about RapMap, an alignment/mapping tool for rna-seq reads to a reference genome, the dataset used for their benchmarks is described like this:

To test the accuracy of different mapping and alignment tools in a scenario where we know the true origin of each read, we generated data using the Flux Simulator (Griebel et al., 2012). This synthetic dataset was generated for the human transcriptome from an annotation taken from the ENSEMBL (Cunningham et al., 2015) database consisting of 86090 transcripts corresponding to protein-coding genes. The dataset consists of 48 million 76 bp, paired-end reads. [...] When benchmarking these methods, reads were aligned directly to the transcriptome, rather than to the genome.

I understand that they generated reads using flux simulator, but I don't understand which transcriptome they mapped them against. I can't find any other information in the paper about which annotation they used either.

I am assuming one would need to generate an entire transcriptome in .fa format from the GTF and genome? Or did they generate a transcriptome from only the simulated reads and then mapped the same reads against it?

GENCODE offers GTFs and FASTAs for gene-encoding transcripts here, might that be the kind of thing they used? Would the reads simulated from these GTFs fit to these transcript FASTA files?

Answer 1

Thanks for your interest in RapMap.

At that time we were using flux simulator for simulating read sequence data. We used the genome and gtf file together as an input to flux.

I dug into the scripts and got hold of the gtf link ftp://ftp.ensembl.org/pub/release-80/gtf/homo_sapiens/Homo_sapiens.GRCh38.80.gtf.gz, and the genome file link ftp://ftp.ensembl.org/pub/release-80/fasta/homo_sapiens/dna/Homo_sapiens.GRCh38.dna.toplevel.fa.

We used rsem-prepare-reference to create the reference transcriptome file later in order to run RapMap and other transcriptome based quantification tools. I hope this helps.

PS: I am a coauthor of the RapMap paper.

Answer 2

If they generated reads from the transcriptome then the started with the FASTA version of the ENSEMBL transcriptome and then aligned the reads back to it. For a project like this I don't think that you need the GTF files, since you don't need to know the location of the transcript on the larger genomic reference. If you can't find an ENSEMBL transcriptome then you can use the GTF and the bedtools program to create a FASTA.