<2020-11-15 Peng Zhang>
Translation is a critical step in gene regulation that synthesizes proteins from a given RNA template. The development of the ribosome profiling (riboseq) technique has enabled the measurement of translation at a genome-wide level. The basic idea of ribosome profiling is to perform deep-sequencing of the ribosome-protected mRNA fragment (~30 nts), termed ribosome footprints, to determine the occupancy of translating ribosomes on a given mRNA. There are several variants of the ribosome profiling technique that are based on the use of different translation inhibitors. The regular ribo-seq utilizes Cycloheximide (CHX), a translation elongation inhibitor to freeze all translating ribosomes. In contrast to CHX, the translation inhibitor lactimidomycin (LTM) and harringtonine (Harr) have a much stronger effect on initiating ribosomes. The use of these two inhibitors allows for the global mapping of translating initiating sites (TISs) when they are coupled with with ribosome profiling (TI-Seq). In addition, when LTM is used sequentially with puromycin (PMY), the TISs can be mapped quantitatively and can be compared between different conditions. we present a novel algorithm, named Ribo TIS Hunter (Ribo-TISH), for identifying translation activities using ribosome profiling data. Ribo-TISH uses statistical tests to assess the significance of translation activities. It captures significant TISs using negative binomial test, and frame biased open reading frames (ORFs) using rank sum test. Ribo-TISH can also perform differential analysis between two TI-Seq data.
Please check the file 'INSTALL.rst' in the distribution.
ribotish [-h] [--version] {quality,predict,tisdiff}
| Example for quality control: | ribotish quality -b ltm.bam -g gene.gtf -t |
|---|---|
| Example for prediction: | ribotish predict -t ltm.bam -b chx.bam -g gene.gtf -f genome.fa -o pred.txt |
| Example for differential TIS: | ribotish tisdiff -1 pred1.txt -2 pred2.txt -a qti1.bam -b qti2.bam -g gene.gtf -o diff.txt |
There are 3 functions available as sub-commands.
| quality: | Quality control for riboseq bam data. |
|---|---|
| predict: | Main function to predict ORF/TIS. |
| tisdiff: | Call diffential TIS between two TIS data |
The main input data is in bam file format. For best performance, reads should be trimmed (to ~ 29 nt RPF length) and aligned to genome using end-to-end mode (no soft-clip). Intron splicing is supported. Some attributes are needed such as NM, NH and MD. For STAR, `--outSAMattributes All` should be set. bam file should be sorted and indexed by samtools.
All positions or regions reported by Ribo-TISH are 0 based, half open, same as in bed format.
Quality control of riboseq bam data. This function checks reads distribution around annotated protein coding regions on user provided transcripts, show frame bias and estimate P-site offset for different group of reads. Reads are grouped by read length as well as 5' end match or mismatch. 5' end mismatch ('m0') reads often have different distribution from matched reads. To turn off 5' end mismatch grouping, use `--nom0`.
There are 3 output files: a txt file recording all distribution data, a pdf figure file and a python file for P-site offset parameters.
Quick examples:
For regular riboseq
ribotish quality -b chx.bam -g gene.gtf
For TI-Seq data
ribotish quality -b ltm.bam -g gene.gtf -t
Riboseq bam data file. Reads should be trimmed and aligned to genome.
Gene annotation file. Acceptable formats include gtf, gff, bed and genepred with gene names. Input file format can be auto detected or specified by `--geneformat` option
Output all distribution data. Default: bampath[:-4]+'_qual.txt'. Quality and offset estimation is based on this distribution. User can save this file for further quick estimation trying different thresholds by `-i` option.
This data is TIS enriched, for LTM and Harr. Quality will pay more attention to TIS sites.
Input previous output file, do not read gene file and bam file again.
Gene annotation file format (gtf, bed, gpd, gff, default: auto)
Input chromosome id mapping table file if annotation chr ids are not the same as chr ids in bam/fasta files. Format:
| chr_name1 | chr_name2 |
Two columns, tab seperated, no specific order requirement. Mappings such as 'chr1' to '1' can be automatically processed without using this option.
Output pdf figure file. Default: bampath[:-4]+'_qual.pdf'
Output offset parameter file. Default: bampath+'.para.py'. This file saves P-site offsets for different reads lengths in python code dict format, and can be used in further analysis.
Range of tag length Default: 25,35. The last number (35) is not included, i.e. the longest length considered is 34.
Position range near start codon or stop codon. Default: -40,20
Number of bins for cds profile. Default: 20
Not consider reads with mismatch at position 0 (5' end mismatch) as a new group.
Threshold for quality. Default: 0.5. Group that frame bias ratio < TH will be considered as low quality and this group of reads will not be used in further analysis. The offset for low quality groups will not be set in parameter file.
Plot RPF 3' end profile instead of 5' end.
Use a color style readable for color blind people ('#F00078,#00F000,#0078F0')
User specified Matplotlib acceptable color codes for three frames (default: 'r,g,b')
Number of processes. Default: 1
Increase output verbosity.
OUTPUT is a txt file recording all distribution data in python format for each group of reads. These distributions are shown in pdf figure file. Quality and offset estimation is based on this distribution. User can save this file for further quick estimation trying different thresholds by `-i` option.
Pdf figure file is plot of all the distributions and illustration of quality and P-site offset. The left part is for 5' end matched reads and the right part is for 5' end mismatch reads if `--nom0` is not set.
Upper panel: the length distribution of RPFs uniquely mapped to annotated protein-coding regions.
Lower panel: different quality metrics for RPFs uniquely mapped to annotated protein-coding regions. Each row shows the RPFs with different lengths.
- Column 1: distribution of RPF 5’ end in 3 frames in all annotated codons. The percentage of the reads from the dominant reading frame is shown.
- Column 2: the distribution of RPF 5’end count near annotated TIS. The estimate of the P site offset and TIS accuracy are also shown. The RPFs of a specific length that do not pass threshold are considered as low quality and removed.
- Column 3: the distribution of RPF 5’end count near annotated stop codon.
- Column 4: The RPF profile throughout the protein-coding regions in 3 frames. TIS enrich score (TIS count / CDS average) is also shown for TIS data.
This file saves P-site offsets for different reads lengths in python code dict format, and can be used in further analysis. The default offset file name is bampath+'.para.py' accompanied with the input bam file. The file format is like
offdict = {28: 12, 29: 12, 30: 12, 32: 13, 'm0': {29: 12, 30: 12, 31: 13}}
The offset parameter file is easy to interpret and can be edited by user if auto estimated offsets are not correct. The default file name will be auto-recognized in further analysis. If the bam file is in a different directory and user do not want to create a parameter file in that directory, we recommend creating a link for the bam file in current working directory, e.g. `ln -s original/dir/ribo.bam`
Ribo-TISH does not guarantee that it can always find best P-site offset values. Users should check the quality figures and edit the parameter file if necessary.
This is the main function of Ribo-TISH. This function predicts ORF/TIS with riboseq bam files. This function uses negative binomial model to fit TI-Seq background and test significance of TIS sites. For regular riboseq data, Wilcoxon rank sum test between in-frame reads and out-frame reads inside the ORF is performed.
Quick examples:
Combine TI-Seq and regular riboseq data
ribotish predict -t ltm.bam -b chx.bam -g gene.gtf -f genome.fa -o pred.txt
For TI-Seq data only
ribotish predict -t ltm.bam -g gene.gtf -f genome.fa -o pred.txt
De novo ORF prediction with only regular riboseq data using longest strategy
ribotish predict -b chx.bam -g gene.gtf -f genome.fa --longest -o pred.txt
De novo ORF prediction with two regular riboseq data using framebest strategy
ribotish predict -b chx1.bam,chx2.bam -g gene.gtf -f genome.fa --framebest -o pred.txt
Only test user provided ORF candidates with two regular riboseq data
ribotish predict -b chx1.bam,chx2.bam -g gene.gtf -f genome.fa -i cand.txt -o pred.txt
Input TI-seq bam data files, comma seperated.
Regular riboseq bam data files, comma seperated.
At least one bam file should be provided by either `-t` or `-b`.
Gene annotation file for ORF prediction. Acceptable formats include gtf, gff, bed and genepred with gene names. Input file format can be auto detected or specified by `--geneformat` option.
If user need to predict on only non-coding genes and use a different gene annotation file for known ORF annotation and background estimation, use `-a` option to provide another gene annotation for known ORF annotation.
If user provided candidates `-i` option is set, the transcript annotation for the candidates should be found in gene annotation file.
Genome fasta file. The fasta file should has a .fai index file accompanied with genome fasta file (indexed) or indexable (fasta sequences have fixed length in each line). This program will index the genome file before prediction if .fai index file can not be found.
Output all possible ORF results that fit the thresholds.
Only test input candidate ORFs, format:
| transID | start | stop |
Start and stop are relative positions to transcript full length and are 0 based, half open. The position 0 is the first base of 5'UTR. Stop - start should be multiples of 3. Use `--igenomepos` if your start and stop are genome positions. Transcript should be found in gene annotation file.
The positions in `-i` input file are genome positions.
Gene annotation file format (gtf, bed, gpd, gff, default: auto)
Input chromosome id mapping table file if annotation chr ids are not same as chr ids in bam/fasta files. See --chrmap option in `quality` section.
Input P-site offset parameter files for `-t` bam files. The default parameter files are bampath+'.para.py' for each bam file, which is generated in `ribotish quality` function. There's no need to specify this option if default parameter files exist. To use this option to provide other parameter files, each bam file should be provided with a file, and file names are separated with comma. If no parameter file is found, default offset 12 will apply for all reads in the bam data.
Input P-site offset parameter files for `-b` bam files. Same as `--tispara` option.
Group transcript according to TIS reads density quantile. Default: 10.
TIS background estimation uses ORF in-frame read counts (excluding TIS codons) to estimate negative binomial parameters. Since different transcripts have different expression levels, the background is different for highly expressed and lowly expressed transcripts. Ribo-TISH groups expressed transcripts into N parts based on TIS reads density of the transcript. Each transcript group have same total number of TIS reads.
Output TIS background estimation result. If only one bam file is provided by `-t` option, the default file name is tisbampath+'.bgest.txt'. If multiple TIS data provided, the default file name is tisBackground.txt
The result file contains negative binomial parameters, group levels and thresholds for each group.
Input background estimation result file instead of instant estimation. By default, if only one bam file is provided by `-t` option, the program will first look for file name tisbampath+'.bgest.txt'. If this file exists, background parameters in this file will be used. Otherwise, TIS background estimation will run and generate a result file according to `-e` option.
Another gene annotation file for ORF annotation in addition to `-g` gene file. This option is mainly used when `-g` annotation focuses on predicting ORFs in non-coding transcripts and does not have sufficient protein coding gene annotation. Protein coding gene annotation is used for TIS background estimation as well as output TIS type classification.
Use alternative start codons. If set, all codons with 1 base different from ATG will be considered as start codon in ORF finding. Affect both TIS background estimation and prediction. It does not affect `-i` mode prediction. To customize alt start codons, use `--altcodons`.
Use provided alternative start codons, comma seperated, e.g. `--altcodons CTG,GTG,ACG`. Turn on `--alt` option. Do not need to provide 'ATG'. It does not support 'N' bases.
Add TIS bam counts to regular riboseq counts. Use TIS data also for ORF frame test. This option will be turned on automatically if `-b` is not provided.
The data is treated with harringtonine (instead of LTM). For Harr data, the reads at TIS sites are not as focus as LTM reads. Reads in flanking region (default 15 codons) of TIS will not be used for TIS background estimation. To customize flanking size, use `--harrwidth`.
Flanking region for harr data, in codons. Default: 15. Turn on `--harr` option.
Only report longest possible ORF results for multiple candidate start codons in the same ORF (same stop codon). This is a TIS selection strategy when there's no `-t` TI-Seq data input.
Only report best frame test results for multiple candidate start codons in the same ORF (same stop codon), which is TIS with the smallest frame test p-value (marked as 'T' in RiboPStatus column). This is a TIS selection strategy when there's no `-t` TI-Seq data input.
Use enrich test instead of frame test. Enrich test is rank sum test between in-frame reads inside ORF and same frame reads outside ORF.
Not require reads compatible with transcript splice junctions.
Minimum amino acid length of candidate ORF, Default: 6.
Only process given genes. Comma separated.
TIS p value threshold. Default: 0.05.
Frame p value threshold. Default: 0.05.
At least one of TIS or frame p value should be lower than this threshold. Default: 1.
Fisher's p value threshold. Default: 0.05.
Fisher's FDR q value threshold. Default: 0.05.
Write all result output without FDR q-value threshold to another file. (default: output + '_all.txt', 'off' or using `--fsqth 1` to turn off)
Number of processes. Default: 1
Increase output verbosity.
Output RPF P-site profile for each transcript. The profile data is in python dict format, recording non-zero read counts at different positions on transcript.
Input RPF P-site profile for each transcript, instead of reading bam reads. The profile file is the output file from `--transprofile` option. Save some time for re-running.
Report ORF sequences.
Report amino acid sequences.
Report all exon block positions for predicted ORFs. Format: start1-stop1,start2-stop2,...startN-stopN. In chromosome direction.
Report the sum of all counts at the in-frame positions in the ORF.
The output is a txt file all possible ORF results that fit the thresholds. Some of the columns are:
| GenomePos: | Genome position and strand of TIS site, 0 based, half open |
|---|---|
| Start: | TIS of the ORF on transcript |
| Stop: | 3' end of stop codon on transcript |
| TisType: | Relative position of this TIS to annotated ORF of the transcript. 'Novel' if no ORF annotation. ':Known' means the TIS is annotated in another transcript. ':CDSFrameOverlap' means the ORF overlaps with annotated CDS in another transcript in the same reading frame. |
| TISGroup: | Group of the transcript for TIS background estimation |
| TISCount: | Number of reads with P-site at TIS site |
| TISPvalue: | One tailed negative binomial test p-value for TISCount (TIS test) |
| RiboPvalue: | One tailed rank sum test p-value for regular riboseq frame bias inside ORF (frame test) |
| RiboPStatus: | For all ORFs sharing same stop codon, 'T' means top (best) p-value, 'L' means local best p-value, 'N' means other. All 'N' in `-i` or `--longest` mode. |
| FisherPvalue: | Combination of TIS and Ribo p-values using Fisher's method |
| TISQvalue: | BH correction q-value of TIS test |
| RiboQvalue: | BH correction q-value of frame test |
| FisherQvalue: | BH correction q-value of Fisher's p-value |
| AALen: | Amino acid length of the ORF |
The '_all' output result is generated according to `--allresult` option, which is similar to the output but do not use FDR (q-value) cutoff. Other cutoffs are the same as output file.
This is the function for differential TIS identification. This function uses two different TIS test results generated by `ribotish predict` using different quantitative TI-Seq (QTI-Seq) data. The ordinary global TI-Seq (GTI-Seq) may have some biases so is not suitable for differential analysis.
First a normalization factor is estimated by Trimmed Mean of M values (TMM) method on the union of significant TIS counts in the two results. Then binomial test p-value and fold change are calculated. If RNASeq counts are provided as reference, the TI efficiency is calculated using Fisher's exact test with normalized count values.
Quick examples:
Differential TIS activity calling
ribotish tisdiff -1 pred1.txt -2 pred2.txt -a qti1.bam -b qti2.bam -g gene.gtf -o diff.txt
Differential TIS efficiency calling with RNASeq count input
ribotish tisdiff -1 pred1.txt -2 pred2.txt -a qti1.bam -b qti2.bam -g gene.gtf --rnaseq RNA.txt -o diff.txt
Predict result of group 1 & 2 TIS data. Comma seperated if there are more than 1 replicates.
Group 1 & 2 TIS riboseq bam files, comma seperated.
Labels for each replicate.
Gene annotation file. Acceptable formats include gtf, gff, bed and genepred with gene names. Input file format can be auto detected or specified by `--geneformat` option.
Output result file.
Gene annotation file format (gtf, bed, gpd, gff, default: auto)
Input P-site offset parameter files for group 1 & 2 bam files. The default parameter files are bampath+'.para.py' for each bam file, which is generated in `ribotish quality` function. To use this option, each bam file should be provided with a file, and file names are separated with comma. If no parameter file is found, default offset 12 will apply for all reads in the bam data.
Not require reads compatible with transcript splice junctions.
Use common TISs instead of union TISs for normalization.
Use only annotated TISs for normalization.
RNASeq count input. Format:
| ID | count1 | count2 | ... |
Both gene ID and transcript ID are acceptable.
Input TIS scale factor of group 2/1 instead of auto calculate. Not log value.
Input RNASeq scale factor of group 2/1 instead of auto calculate. Not log value.
Export count table for differential analysis with other tools. Especially for replicated data.
Scatter plot output pdf file.
Scatter plot figure size. Default: 8,8.
Minimum fold change threshold. Default: 1.5.
Input TIS p value threshold. Default: 0.05.
Input TIS q value threshold. Default: 0.05.
Output TIS diff p value threshold. Default: 0.05.
Output TIS diff q value threshold. Default: 0.05.
Number of processes. Default: 1
Increase output verbosity.
The output is a txt file all differential TIS results that fit the thresholds. Some of the columns are:
| FoldChange: | Fold change (2/1) value after normalization |
|---|---|
| DiffPvalue: | Differential test p-value, two-tailed. |
| DiffQvalue: | BH correction q-value of DiffPvalue |
The export table is generated using `--export` option. It is also automatically generated when the input data has replicated samples. It is a txt file with raw TIS counts for each predicted TIS. The format of TIS id is 'TransID_Start_GenomePos'.
For replicated data, Ribo-TISH provided R scripts to call differential TISs using edgeR or DESeq2.
Example for edgeR:
Rscript path_to_scripts/tisdiff_edgeR.r tisdiff_export.txt 3 4 tisdiff_edgeR_output.txt
For DESeq2:
Rscript path_to_scripts/tisdiff_DESeq2.r tisdiff_export.txt 3 4 tisdiff_DESeq2_output.txt
3 and 4 are number of replicates in two conditions.
If `--rnaseq` is provided, the RNASeq counts of genes/transcripts for the TISs are also provided in the export table. However, the analysis for RNASeq referenced differential TIS efficiency analysis with replicate data is currently unavailable.