Introduction

Welcome to RiboCrypt

RiboCrypt is an R package and web server for interactive visualization and analysis in genomics. RiboCrypt works with most NGS-based methods, but much emphasis is put on Ribo-seq data visualization. RiboCrypt uses ORFik experiment objects, so to ensure compatibility use ORFik and massiveNGSpipe for processing own (unpublished) data. If you encounter any issues, please, contact us using the info in the footnote.

The following is web server walk through listing all utilities and options therein.

Ribo-seq overview

For more general material about ribosome profiling, we suggest reading this review. For more technical explanation of data processing steps and specific analyses, it’s best to see ORFik Overview vignette, especially chapter 6 “RiboSeq footprints automatic shift detection and shifting”.

Browser

The browser is the basic coverage display page. It contains a selection panel on the left side and a display window (browser) on the right. It displays coverage of sequencing data in either transcript coordinates (default, collapsed introns), or genomic coordinates. The following (Fig. 1), SRD5A1 gene, has recently been shown to be decoded in three frames.

Figure 1. SRD5A1 gene displayed using default options. Using API, this browser view can be re-generated using this link.

Browser window

When you press the “plot” button on the selection panel, the selected data will be displayed according to specified options, explained later in the tutorial.

The browser window consists of the specific parts:

Single or multiple density tracks are displayed on top. By default Ribo-seq is rendered in 3 colors, where

red is 0 frame, the start frame of reference transcript.
green is +1 frame
blue is +2 frame

Sequence track (top middle), displayes DNA sequence when zoomed in (< 100nt)
Annotation track (middle), the annotation track displays the transcript annotation, together with black bars that is displayed on top of the data track. CDS and other annotations are color-coded according to relative reading frame
Frame track (bottom), the 3 frames displayed with given color bars:

white (start codon, ATG)
black (stop codons, TAA | TAG | TGA)
brown (custom motifs, e.g. NTG all start codon alternatives)

When zoomed in, the amino acid sequence is displayed within each frame.

Additional displays, for example protein structure viewer, which appears after clicking on isoform ID on annotation track.

In the example above (Fig. 1), notice green coverage preceding CDS (in Blue). It corresponds to overlapping open reading frame, so it is clear that there are two regions that undergo translation simultaneously in this locus.

Selection panel (browser)

The display panel shows the primary settings, (study, gene, sample, etc):

Experiment

Select an organism: Either select “ALL” to keep all experiments, or select a specific organism to select display only that subset of experiments in experiment select tab.
Select an experiment: The experiments contain study names combined with organism (some studies are multi species, so sometimes one study have multiple experiments). Select which one you want. There also exist merged experiments (all samples merged for the organism, etc). RNA-seq experiments have an addition “_RNA” as suffix.

Gene

Select a gene: A gene can be selected using:
- Gene id (ENSEMBL)
- Gene symbol (HGNC, etc)
Select a transcript: A transcript isoform of the given gene above, default is Ensembl canonical isoform, other isoforms can be selected.

Library

Usually each experiment have multiple libraries. In the case of merged experiments, different modalities (RFP, RNA, disome, TI, see Fig. 2) can be selected in this field. Select which one to display, if you select multiple libraries they will be stacked in the browser as multiple tracks.

Library are by default named:

Library type (RFP, RNA etc),
Condition (WT, KO (wild type, knock out ) etc)
Stage/timepoint (5h, 1d (5 hours, 1 day) etc)
fraction (chx, cytosolic, ATF4 (ribosomal inhibitor, cell fraction, gene) etc)
replicate (technical/biological replicate number (r1, r2, r3))

The resuting name could for example be:

RFP_WT_5h_chx_cytosolic_r1

It’s normal to see that if condition is KO (knockout), the fraction column contains a gene name (the name of the gene that was knocked out). Currently the best way to find SRR run number for respective sample is to go to metadata tab and search for the study.

Display type

Select reading frames display type:
- lines (single line per frame, most clear for middle distance (> 100 nt))
- columns (single bars per position, most clear for single nucleotide resolution)
- stacks (Area under curve, stacked, most clear for large regions (> 1000 nt))
- area (Area under curve, with semitransparent overlapping frames, most clear for large regions (> 1000 nt))
- heatmap (good for displaying large number of libraries, see metabrowser for display of thousands of libraries)
K-mer length: sliding window of selected length applying averaging coverage per frame. K-mer = 1 means unaltered data. When looking at a large region (> 100nt), pure coverage can usually be hard to inspect due to intrinsic Ribo-seq spikyness. Using K-mer length > 1 (9, the default, is a good starting point), you can easily look at patterns over larger regions.

Notice, how in figure 2 reducing K-mer length to 1, changing display type to columns and adding TI-seq (Translation Initiation profiling, samples treated with translation initiation inhibitors - either lactimidomycin or harringtonin) enables detection of translation in all three reading frames. In fact, experienced user can notice drop of signal in red frame after stop codon of the corresponding ORF (hidden behind most abundant, green), further corroborating overlapping translation. If this process seems somewhat imprecise and not fully defined - it is by design - annotating such elusive phenomena is notoriously difficult to tackle algorithmically, and doing so successfully requires deep intuition in how the signal and noise behave, so manual investigation into many instances like this. Thus, RiboCrypt can be treated as a hypothesis building tool, rather a database of fixed answers. Try modifying yourself, for example switching on RNA-seq track, or altering smoothing window width.

Figure 2. Comparison of display types and smoothing window widths.

Selection panel (Settings)

Settings tab contains additional parameters:

5’ extension (extend viewed window upstream)
3’ extension (extend viewed window downstream)
Custom sequences highlight (Motif search, given in brown color, support IUPAC, examples: CTG or NTG)
Genomic region (Browse genomic window instead of gene, syntax: chromosome:start-stop:strand, human/mouse/zebrafish: 1:10000-20000:+ , yeast: I:10000-20000:+. Both 1 and chr1 works, conversion will be done automatically)
Zoom interval (start with a zoom and highlight on a subsection): Either specified in tx/view coordinates, i.e. 20:50 will give zoom on region (20-10):(50+10) = 10:60, and a highlight color (light yellow) at coordinates 20:50. You can also specify in genomic coordinates as above for genomic region (remember the genomic coordinates must then be within the gene / region you are displaying).
Genomic View (Activate/deactivate genomic view, displaying full introns, see Figure 3)
Full annotation (display all transcript isoforms annotation or just the selected isoform)
uORF annotation (display candidate uORFs in the annotation track) this is not the predicted set, but all possible uORFs by seqnece. For predicted set, check the predicted translons checkbox.
Predicted translons - display predicted translons in the annotation track. Try turning it on for the example in Figures 1 and 2 - notice that the red (second) translon was missed
Log scale (Log scale the coverage, reduces effect of extreme peaks)
Protein structures (If you click the annotation name of a transcript in the plot panel it will display the alpha-fold protein colored by the ribo-seq data displayed in the plot panel)
PhyloP (PhyloP conservation track will be added on the bottom)
Split color frame (Split riboseq signal into colors by frame) i.e. red, green and blue. If this is on, then Ribo-seq and Ti-seq will be displayed with colors split by frame.
Frames subset (If split color frame is TRUE, you can display only wanted color)
Summary top track (Add an additional plot track on top, summarizing all selected libraries)
Select Summary display type (same as frames display type above, but for the summary track)
Export format (Choose plot export format from plotly controls. The default is svg vector graphics that allows for high customizability)
Plot (red button, render selected gene/region as plotly html, with specified settings)

Uniquely, RiboCrypt allows for toggling between transcriptomic and genomic views, while displaying coding exons in colors according to the correct reading frame. Moreover, extensions allow for exploration of large chunks of the genome, even tens of thousands of bases. Try finding all genes in this browser window, and then turn full annotation on to see what you missed! (loading may be a bit slow, but displaying over region 30 thousands bases long here isn’t practical)

Figure 3. Comparison between transcriptomic and genomic view on yeast EFM5 and ABP140 genes.

Mega Browser

Display all samples for a specific organism over selected gene (Fig. 5). This tab does not use bigwig files to load (as that would be very slow). It uses precomputed fst files of coverage over all libraries. Note: Not all isoforms are computed, by default the longest isoform is computed.

Figure 5. Heatmap of thousands of libraries coverage over ATF4 transcript. Clustered using k-means = 2, with summary track displayed on top.

Selection panel (Browser)

Organism, experiment and gene explained above

Group on: the metadata column to order plot by
K-means clusters: How many k-means clusters to use, if > 1, Group will be sorted within the clusters, but K-means have priority.

Selection panel (Settings)

Normalization (all scores are tpm normalized and log scaled)
- transcriptNormalized (each sample counts sum to 1) (default)
- Max normalized (each position count divided by maximum)
- zscore (zscore normalization, (count * mean / sd) variance scaled normalization)
- tpm (raw tpm of counts, is very sensitive to extreme peaks)
Color theme: Which color theme to use
Color scale multiplier: how much to amplify the color signal (if all is single color, try to reduce or increase this depending on which color is the majority)
Sort by other gene: Sort heatmap on expression of another gene (increasing). A line plot will be shown left that display the expression of that other gene per row. Also the enrichment analysis is done on bins of the other genes expression instead of metadata term.
Summary top track (same as in browser, the aggregate of all rows, useful to see the frame information, which is not represented in the heatmap)
Split by frame: (Not active yet!): When completed will display the heatmap with frame information, so rgb colors from white to dark red etc.
Summary display type: Same as in browser
Region to view: By default uses entire transcript. Can subset to only view cds or leader+cds etc.
Select a transcript: Which transcript isoforms not all isoforms have computed fsts. By default the longest isoforms have been computed for all genes.
K-mer length: Smoothens out the signal by applying a mean sliding window, default 1 (off)

Statistics tab

This tab gives the statistics of over representation analysis per cluster of the metabrowser plot. Using chi squared test, it gives the residuals per term from metadata (like tissue, cell-line etc). If a value is bigger than +/- 3, it means it is quite certain this is over represented. This is shown as a red line.

If no clustering was applied, this tab gives the number of items per metadata term (40 brain samples, 30 kidney samples etc).

Requirements

This mode is very intensive on CPU, so it requires certain pre-computed results for the back end. That is namely: - Premade collection experiments (an ORFik experiment of all experiments per organism) - Premade collection count table and library sizes (for normalizations purpose) - Premade fst serialized coverage calculation per gene (for instant loading of coverage over thousands of libraries)

Note that on the live app, the human collection (4000 Ribo-seq samples) takes around 30 seconds to plot for a ~ 2K nucleotides gene, ~99% of the time is spent on rendering the plot, not actual computation. Future investigation into optimization will be done.

Analysis

Here we collect the analysis possibilities, which are usually operating on meta-gene or multi locus scale.

Heatmap

This tab displays a heatmap of coverage per readlength at a specific region (like start site of coding sequences) over all genes selected.

Figure 4. Metagene per-readlength heatmap of before (upper panel) and after (bottom panel) P-shifting. Notice emerging periodicity.

Selection panel (heatmap)

Study and gene select works same as for browser specified above. In addition to have the option to specify all genes (default).

Select libraries. Currently Only 1 library can be selected in heatmap mode.
View region Select one of:
- Start codon
- Stop codon
Normalization Normalization mode for data display, select one of:
- transcriptNormalized (each gene counts sum to 1)
- zscore (zscore normalization, will give better overview if 1 readlength is extreme)
- sum (raw sum of counts, is very sensitive to extreme peaks)
- log10sum (log10 sum of counts, is less sensitive to extreme peaks)
Min Readlength The minimum readlength to display
Max readlength The maximum readlength to display

Selection panel (settings)

Here additional options are shown:

5’ extension (extend viewed window upstream from point, default 30)
3’ extension (extend viewed window downstreamfrom point, default 30)
Summary top track - summarized coverage from all readlengths
p-shifted - display either P-sites (default), or reads 5’ ends.

Metagene analysis with heatmap module can be used, for example, to investigate how well P-site positioning was performed (Fig. 4)

Codon analysis

This tab displays a heatmap of codons dwell times over all genes selected, for both A and P sites. When pressing “Differential” you swap to a between library differential codon dwell time comparison (minimum 2 libraries selected is required for this method!)

Display panel (codon)

Study and gene select works same as for browser specified above. In addition to have the option to specify all genes (default).

Select libraries (multiple allowed)

Filters

Codon filter value (Minimum reads in ORF to be included)
Codon score, all scores are normalized for both codon and count per gene level (except for sum):
- percentage (percentage use relative to max codon, transcript normalized percentages)
- dispersion(NB) (negative binomial dispersion values)
- alpha(DMN) (Dirichlet-multinomial distribution alpha parameter)
- sum (raw sum, (a very biased estimator, since some codons are used much more than others!))

Codon plot

Display is the score per codon (amino acid), in addition there are 2 custom “amino acids”, * as in * : TGA, means TGA is a stop codon (last codon in CDS). Similar is #, as in # : ATG which means ATG as start codon (first codon in CDS). For P-sites start codons should be enriched, while for A-sites there should be a richer variability, often with a small enrichment for stop-codons. We will implement a richer model eventually using the more correct negative binomial relationships between E, P and A sites, i.g. the motif PPP (triplet-proline in E,P,A site) is much stronger than a single P in the A site etc. alone etc.

Differential gene expression

Given an experiment with a least 1 design column with two values, like wild-type (WT) vs knock out (of a specific gene), you can run differential expression of genes. The output is an interactive plot, where you can also search for you target genes, making it more useable than normal expression plots, which often are very hard to read.

Selection panel (Differential expression)

Organism and experiment explained above - Differential method: FPKM ratio is a pure FPKM ratio calculation without factor normalization (like batch effects), fast and crude check. DESeq2 argument gives a robust version, but only works for experiments with valid experimental design (i.e. design matrix must be full ranked, see deseq2 tutorial for details!) - Select two conditions (which 2 factors to group by)

Selection panel (settings)

draw unregulated (show dots for unregulated genes, makes it much slower!)
Full annotation (all transcript isoforms, default is primary isoform only!)
P-value (sliding bar for p-value cutoff, default 0.05)
export format for plot (explained above)

Read length (QC)

This tab displays a QC of pshifted coverage per readlength (like start site of coding sequences) over all genes selected.

Display panel (Read length QC)

The display panel shows what can be specified to display, the possible select boxes are same as for heatmap above:

Plot panel

From the options specified in the display panel, when you press “plot” the data will be displayed. It contains the specific parts:

Top plot: Read length relative usage

Y-axis: Score
Color: Per frame (red, green, blue)
Facet box: the read length

Bottom plot: Fourier transform (3 nucleotide periodicity quality, clean peak means good periodicity)

Fastq (QC)

This tab displays the fastq QC output from fastp, as a html page.

Display panel (Read length QC)

The display panel shows what can be specified to display, you can select from organism, study and library.

Plot panel

Displays the html page.

Metadata

Metadata tab displays information about studies and custom predictions. RiboCrypt is integrated with Ribo-seq Data Portal, refer to this paper for details on metadata curation and standardization.

## SRA search

Search SRA for full information of supported study

Study accession number

Here you input a study accession number in the form of either:

SRP
GEO (GSE)
PRJNA (PRJ….)
PRJID (Only numbers)

Output

On top the abstract of the study is displayed, and on bottom a table of all metadata found from the study is displayed.

Studies

Full table of supported studies with information about sample counts

Predicted translons

Full list of predicted translons on all_merged tracks per species. Each translons has a link that directs you to the browser tab and displays the transcript zoomed in on the translons with a light yellow highlighting of the region.

Settings:

Translon annotation scheme:

Prediction algorithm: ORFik::detect_ribo_orfs
translon_types <- c(“uORF”, “uoORF”,“doORF”,“dORF”)
start_codons <- “ATG|TTG|CTG”
additional arguments: add sequence and amino acid sequence per translon
All other arguments default.

Additional information

All files are packed into ORFik experiments for easy access through the ORFik backend package:

File formats used internally in experiments are:

Annotation (gtf + TxDb for random access)
Fasta genome (.fasta, + index for random access)
Sequencing libraries (all duplicated reads are collapsed)
- random access (only for collapsed read lengths): bigwig
- Full genome coverage (only for collapsed read lengths): covRLE
- Full genome coverage (split by read lengths): covRLElist
count Tables (Summarized experiments, r data serialized .rds)
Library size list (Integer vector, .rds)
Precomputed gene coverages per organism: fst (used for metabrowser)

massiveNGSpipe

For our webpage the processing pipeline used is massiveNGSpipe which wraps multiple tools:

Fastq files are download with ORFik download.sra
Adapter is detected with either fastqc (sequence detection) and falls back to fastp auto detection.
Reads are then trimmed with fastp (using the wrapper in ORFik)

Adapter removal specified
minimum read size (20nt)

Read are collapsed (get the set of unique reads and put duplication count in read header)
Reads are aligned with the STAR aligner (using the wrapper in ORFik), that supports contamination removal. Settings:

genomic coordinates (to allow both genomic and transcriptomic coordinates)
local alignment (to remove unknown flank effects)
minimum read size (20nt)

When all samples of study are aligned, an ORFik experiment is created that connects each sample to metadata (condition, inhibitor, fraction, replicate etc)
Bam files are then converted to ORFik ofst format
These ofst files are then pshifted
Faster formats are then created (bigwig, fst and covRLE) for faster visualization

About

This app is created as a collaboration with:

University of Warsaw, Poland
University of Bergen, Norway

Main authors and contact:

Michal Swirski (Warsaw), email: [email protected]
Håkon Tjeldnes (Bergen), email: [email protected]

App Tutorial

Introduction

Ribo-seq overview

Browser

Browser window

Selection panel (browser)

Experiment

Gene

Library

Display type

Selection panel (Settings)

Mega Browser

Selection panel (Browser)

Selection panel (Settings)

Statistics tab

Requirements

Analysis

Heatmap

Selection panel (heatmap)

Selection panel (settings)

Codon analysis

Display panel (codon)

Filters

Codon plot

Differential gene expression

Selection panel (Differential expression)

Selection panel (settings)

Read length (QC)

Display panel (Read length QC)

Plot panel

Fastq (QC)

Display panel (Read length QC)

Plot panel

Metadata

Study accession number

Output

Studies

Predicted translons

Settings:

Additional information

massiveNGSpipe

API for URL access and sharing

Primary url:

Page selection API:

Parameter API:

browser:

About