RDG: Ribosome Decision Graph

Documentation & User Guide

Overview

The Ribosome Decision Graph (RDG) is an interactive visualization tool for exploring alternative translation initiation in eukaryotic mRNAs. It models how scanning ribosomes encounter multiple start codons and make probabilistic decisions to initiate or continue scanning, ultimately producing diverse protein products (translons) from a single mRNA sequence.

What is a Ribosome Decision Graph?

The RDG is a directed graph that represents all possible translation paths through an mRNA:

Nodes: Decision points at each start codon where ribosomes choose to initiate or scan past
Edges: Translation events (translons) from start to stop, or scanning events moving to the next decision point
Probabilities: Each path has an associated probability based on start codon context (Kozak sequence, codon type, local structure)
Outcomes: Final protein products with predicted relative abundances

Getting Started

Loading Sequences

You can load sequences in three ways:

Manual Input: Paste any mRNA sequence (AUGC format) directly into the text area
Gene Symbol: Enter a human gene symbol (e.g., ATF4, BRCA1, TP53) to fetch the canonical transcript from Ensembl
Transcript ID: Enter a specific Ensembl transcript ID (e.g., ENST00000337304) for precise control

Note: Fetched sequences are trimmed to the first 1500 nucleotides for optimal performance. This typically includes the full 5'UTR and initial coding region where most regulatory elements reside.

Understanding the Visualization

The RDG has two main view modes:

Tree View: Shows the complete decision tree with branching paths at each start codon
- Frame tracks (green, blue, orange) represent the three reading frames
- Circles indicate start codon positions (size = Kozak strength)
- Bars show translation events from start to stop
- Branch points show where ribosomes decide: initiate (go down) or scan past (continue right)
- Edge labels display probabilities and translon names (e.g., "T1 (45.2%)")
Flux View: Shows ribosome traffic through the sequence over time
- Animated ribosomes (colored circles) scan 5' to 3'
- Ribosome color indicates which frame they're aligned to
- Watch how ribosomes accumulate at strong start codons and how some scan past weak ones
- Use simulation speed controls (1x, 2x, 5x, 10x) to adjust animation

Working with Translons

Adding Translation Events

Translons are added automatically based on detected start codons, but you can also add them manually:

Click anywhere on the canvas to add a translon starting at that position
The tool finds the nearest downstream stop codon in the same frame
Initial probability is set based on Kozak sequence context
Fine-tune probability using the slider in the translon controls panel

Translon Details

Click on any translon card to expand and view detailed information:

Position: Start and stop coordinates in the mRNA
Frame: Reading frame (0, 1, or 2)
ORF Length: Size of the open reading frame
Start Codon: Actual codon sequence (AUG or near-cognate)
Kozak Context: Flanking sequence and quality score
- Strong (≥0.7): Efficient initiation expected
- Moderate (0.4-0.7): Context-dependent initiation
- Weak (<0.4): Leaky scanning likely
GC Content: Local RNA structure indicator (high GC = potential structure)

Adjusting Probabilities

Each translon has an adjustable initiation probability (0-99%):

Move the slider to change probability
The RDG updates in real-time to reflect your changes
Probabilities affect downstream translons - if 80% initiate at T1, only 20% are available for T2
Use this to test different biological scenarios and parameter sensitivities

Understanding the RDG Model

Translation Probability Calculation

The model calculates initiation probability at each start codon:


            P_init = P_base × f_Kozak × f_codon × f_structure

P_base (Base Probability): Fundamental initiation rate (default 0.3, adjustable in RDG Parameters)
f_Kozak (Kozak Factor): Context strength, 0.5-1.0 based on positions -3 and +4
- Optimal context: gccA/GccAUGG
- Position -3 = A or G: +0.5
- Position +4 = G: +0.5
f_codon (Codon Factor): Codon type efficiency
- AUG: 1.0 (100% efficiency)
- CUG, GUG, ACG, etc.: 0.1 (10% efficiency, adjustable via "Near-Cognate Penalty")
f_structure (Structure Factor): Local secondary structure effect
- Based on GC content in 20nt window
- High GC (>50%) increases probability (structure may pause ribosomes)
- Effect controlled by "GC Bonus" parameter

Adjustable Parameters

Fine-tune the model using the RDG Parameters section:

Base P (0.1-0.5): Baseline initiation probability - increase for generally higher initiation rates
Near-Cognate Penalty (0.05-0.3): Non-AUG start efficiency - higher values make CUG/GUG more competitive
GC Bonus (0.0-1.0): Structure-mediated enhancement - higher values make GC-rich regions favor initiation more

Click "Apply to All Translons" after adjusting parameters to recalculate all probabilities with your new settings.

Reinitiation Model

After translating a short upstream ORF (uORF), ribosomes may reinitiate downstream:

Reinitiation is more likely after short uORFs (<50 codons)
Probability decreases with uORF length (ribosomes lose factors during long translation)
The model shows reinitiation paths as branches continuing from stop codons
This explains how genes like ATF4 and GCN4 achieve stress-dependent upregulation

Predicted Protein Products

The Products panel shows all predicted proteins with:

Name: Translon identifier (T1, T2, etc.)
Length: Protein length in amino acids (ORF nt ÷ 3)
MW: Molecular weight in kDa (assuming 110 Da average per residue)
Abundance: Relative amount (%) based on initiation probability
- Accounts for sequential depletion (if 80% initiate at T1, max 20% available for T2)
- Normalized so total across all products = 100%

Interactive Features

Simulation Controls

Play/Pause: Start/stop the ribosome scanning simulation (flux view)
Clear: Remove all ribosomes from the simulation
Speed Buttons: Adjust animation speed (1x, 2x, 5x, 10x)

Naming & Export

Name This Analysis: Save your current RDG with a descriptive name for reference
Export as Video: Record the canvas as a WebM video file (works best in flux view to capture animation)

Practical Applications

Studying Gene Regulation

uORF-Mediated Regulation: Analyze genes like ATF4, ATF5, or GCN4 to understand how upstream ORFs control downstream translation
Leaky Scanning: Compare strong vs. weak Kozak contexts to predict bypass efficiency
Alternative Isoforms: Identify potential N-terminal variants from downstream start codons
Near-Cognate Initiation: Detect non-AUG start sites that may produce functional proteins

Optimizing Constructs

Reporter Design: Predict how 5'UTR sequences affect downstream reporter expression
Expression Tuning: Test different Kozak sequences to optimize protein production
Background Reduction: Identify and eliminate unintended start codons

Best Practices & Tips

Start Simple: Load a well-characterized gene (e.g., ATF4) to understand the interface before analyzing your own sequences
Focus on 5'UTR: Most alternative initiation happens in the first few hundred nucleotides - trimming sequences to ~1500 nt is usually sufficient
Check Kozak Scores: Strong contexts (>0.7) are efficient initiation sites; weak contexts (<0.4) often permit leaky scanning
Use Both Views: Tree view for understanding decision logic, flux view for visualizing ribosome dynamics
Test Parameters: Adjust base probability and penalties to see how sensitive your predictions are to model assumptions
Compare Variants: Load wild-type sequence, adjust one parameter or remove one uORF, and compare products
Name Your Work: Use descriptive names (e.g., "ATF4 WT", "ATF4 uORF1-mut") when saving analyses

Limitations & Considerations

Simplified Model: Real translation involves many factors not captured here including:
- RNA secondary structure (only approximated via GC content)
- RNA-binding proteins and upstream ORF peptide effects
- Ribosome availability and initiation factor concentrations
- Context-specific regulation (cell type, stress conditions)
No Stop Readthrough: Model assumes all stop codons terminate translation (readthrough/suppression not implemented)
No Frameshifting: Programmed ribosomal frameshifts are not currently supported
Static Probabilities: Initiation probabilities are based on sequence alone, not dynamic cellular conditions
Prediction Tool: Results are hypothetical - experimental validation (e.g., ribosome profiling, western blots) is essential

Example Workflows

Workflow 1: Analyzing ATF4 uORF Regulation

Select "Gene Symbol" and enter "ATF4"
Click "Fetch Sequence" - loads canonical transcript 5'UTR
Switch to Tree View to see the complete RDG
Observe two main uORFs (uORF1 and uORF2) before the main ORF
Note that uORF1 is short - ribosomes likely reinitiate after translating it
uORF2 overlaps with the main ORF start - creates competing initiation
Under normal conditions (adjust base P to ~0.3), uORF2 captures most ribosomes, limiting main ORF translation
Simulate stress (increase base P to ~0.4-0.5) - more ribosomes initiate at uORF1, fewer reach uORF2, more reinitiate at main ORF
This explains ATF4's paradoxical stress-induced upregulation

Workflow 2: Optimizing a Reporter Construct

Paste your 5'UTR + reporter ORF sequence
Review all detected start codons in the sequence
Identify any upstream start codons that might compete with your reporter
Check their Kozak scores - strong contexts (>0.7) may significantly reduce reporter expression
Consider mutating problematic uORFs (change AUG to AAG) or adjusting Kozak context
Use the abundance predictions to estimate reporter yield
For dual luciferase reporters, see the Western Blot demo for protein product predictions

Technical Details

Reading Frame Color Coding

Frame 0 (Green): Positions 0, 3, 6, 9... (divisible by 3)
Frame 1 (Blue): Positions 1, 4, 7, 10... (remainder 1 when divided by 3)
Frame 2 (Orange): Positions 2, 5, 8, 11... (remainder 2 when divided by 3)

Start Codon Detection

The tool automatically detects:

AUG: Standard methionine start codon
Near-cognates: CUG (Leu), GUG (Val), ACG (Thr), AUU (Ile), AUC (Ile), AUA (Ile) - all shown but with reduced efficiency

Abundance Calculation

Product abundance accounts for sequential ribosome depletion:


            Available ribosomes = 100%

            T1 initiates at 45% → produces 45% T1, leaves 55% available

            T2 initiates at 60% of remaining → produces 33% T2 (0.55 × 0.60), leaves 22% available

            T3 initiates at 80% of remaining → produces 17.6% T3 (0.22 × 0.80), leaves 4.4% available

Shared Code Architecture

This tool uses modular JavaScript libraries:

rdg-engine.js: Core calculation engine
- Start codon detection
- Kozak scoring
- Probability calculations
- Stop codon finding
- Protein product prediction
rdg-viz.js: Visualization layer
- Tree layout algorithm
- Canvas rendering
- Construct suggestions

These modules are shared with the Western Blot demo for consistent behavior and easier maintenance.

Related Tools

Western Blot Demo: Use the same RDG model to predict protein products from reporter constructs and visualize them as gel bands
Ribosome Profiling: Compare RDG predictions with experimental ribosome profiling data to validate and refine the model

Glossary

Translon: A translation event producing a specific protein isoform from a particular start-stop codon pair
uORF (upstream ORF): Short open reading frame in the 5'UTR that can regulate downstream translation
Kozak Sequence: Nucleotide context surrounding a start codon that affects initiation efficiency
Leaky Scanning: Process where ribosomes bypass a start codon and continue scanning downstream
Reinitiation: Resumption of scanning and translation after terminating a short uORF
Near-Cognate Codon: Non-AUG codon (e.g., CUG, GUG) that can initiate translation at reduced efficiency
Reading Frame: One of three possible ways to read a sequence as triplet codons (frames 0, 1, 2)
ORF (Open Reading Frame): Sequence from a start codon to a stop codon without internal stops