PROTEOFORMER is a proteogenomic pipeline that delineates true in vivo proteoforms and generates a protein sequence search space for peptide to MS/MS matching. It can be combined with canonical protein databases or used independently for identification of novel translation products. The pipeline makes use of the recently developed next generation sequencing strategy termed ribosome profiling (RIBO-seq) that provides genome-wide information on protein synthesis in vivo. RIBO-seq is based on the deep sequencing of ribosome protected mRNA fragments. RIBO-seq allows for the mapping of the location of translating ribosomes on mRNA with sub codon precision, it can indicate which portion of the genome is actually being translated at the time of the experiment as well as account for sequence variations such as single nucleotide polymorphisms and RNA splicing.

PROTEOFORMER takes as input two fastq files (NGS reads files representing ribosome-protected fragments (RPFs)) of the elongating and initiating ribosomes and outputs a FASTA protein sequence database of derived translation products based on Ensembl transcript annotations. Furthermore, specific metrics (e.g. FPKM, coverage, gene RPF abundance) are deduced to enable the verification of the RIBO-seq data quality. The alignment and RPF density information is outputted in order to allow easy upload and visual evaluation in a genome browser environment.The pipeline consists of eight major parts:

1. Raw Input Quality Control: This preliminary step determines the general quality of the raw sequencing data using FastQC.
2. Mapping: The first step makes use of transcriptome mappers (STAR or TopHAT2) to align RPFs from the input files against a references genome based on the corresponding Ensembl annotation bundle. Before mapping, reads will be preprocessed (adaptor clipping, quality filtering, pre-mapping against rRNA, tRNA,…).
3. Post-mapping quality control: This step analyses the quality of the preprocessed, aligned reads using FastQC and mQC. At the same time, the user gets a general overview of the data outlook.
4. Transcript calling: This step makes use of the ribosome profiles of the elongating ribosomes to mark transcripts with experimental evidence of translation. Extra annotation is also added in this step (CCDS id, canonical transcript).
5. TIS calling: This entails the identification of translation initiation sites (TIS). It implements a rule-based algorithm that combines RIBO-seq information from two related translation inhibitors: an elongation inhibitor (e.g. cycloheximide) and an initiation inhibitor (e.g. lactimidomycin) to differentiate TIS sites from elongating ribosomes.
6. Variation calling: This part of the pipeline uses samTools and/or a dbSNP to identify variants in the mapped reads.
7. Translation assembly: This step assembles all translation products based on the TIS, transcript isoform, and/or SNP information derived from the RIBO-seq data.
8. Translation Database: Finally a (non-)redundant FASTA-formatted database of derived translation products is generated wherein eventually all duplicate and sub-sequences are removed.
9. Floss Calculation: Calculates first the reference fractions and cut off values based on known protein-coding transcripts. With these, the FLOSS scores are calculated and classified for each possible translation product.
10. Feature summarization: The user can generate lists of RPF counts per feature (gene, transcript, exon, ORF, promotor).

Datasets without initiation inhibitor data lanes cannot be analysed with TIS calling, variation calling and translation assembly. Instead, tools are developed which use PRICE or SPECtre to determine the translated open reading frames.
Check our manuscript for full explanation. Also please cite our work if you plan on using it (http://dx.doi.org/10.1093/nar/gku1283). The full user manual can be found on GitHub.

PROTEOFORMER was developed in Perl 5 and is freely available for download in a script version and as a Galaxy implementation.