PETAL tutorial
Use PETAL to extract gene regulatory information from plant DNA.
PETAL helps you screen plant regulatory DNA, inspect promising candidate sequences and modify a sequence's regulatory activity. This guide focuses on what tool to run, what the results mean and how to avoid over-interpreting them.
What do you want to do?
Pick the PETAL tool based on your question!
Predict activity
Use PETAL Predict when you have one or many sequences and want to predict their regulatory activity.
Inspect important bases
Use PETAL Interpret for selected sequences when you want to see which positions drive their activity.
Make a figure
Use PETAL Plot to create prediction profiles or Interpret sequence logos from your PETAL results.
Modify regulatory activity
Start from one 170 bp sequence and let PETAL propose small mutations toward your chosen objective.
A simple PETAL analysis
Prepare your DNA sequence input
Paste sequences or upload a FASTA file. Use only A, C, G and T. Sequence names become the names in the result table.
Choose the correct model
Pick the model that best matches the type of regulatory element and experimental system you care about. The model page explains the available choices.
Run PETAL Predict first
Prediction is the screening step. It gives one activity estimate per condition and per 170-bp window.
Calculate attributions with PETAL Interpret
PETAL Interpret screens your sequence and calculates an attribution score for each nucleotide. From the attributions you can discover patches of high importance for your element's activity.
Plot or inspect the interesting hits
Use PETAL Plot for overview figures. A 170-bp sequence is shown as a bar plot. Longer sequences are scanned with overlapping 170-bp windows in 1-nt steps and shown as a prediction profile. Use PETAL Plot for Interpret data to generate sequence logos highlighting patches of high importance in your sequence.
Use PETAL Evolve to modify regulatory activity
In each round, PETAL Evolve introduces single-nucleotide mutations in your target sequence and keeps the sequence variant most closely matching your objectives for the next round. Evolved sequences are candidates for follow-up experiments. They are not guaranteed to work in plants or in every assay context. They also may still contain unwanted motifs, such as restriction enzyme cut sites!
Remove unwanted motifs with PETAL Domesticate.
Sequences derived from PETAL Evolve (or any other sequences) may still contain unwanted motifs, such as restriction enzyme cut sites. PETAL Domesticate iteratively removes given motifs, trying to minimize the predicted effect on the regulatory activity of the sequence.
Sequence input that works well
FASTA input
>candidate_1 GGAGTTCAGTTCGGTAACGGAGAGACTCTGCGGCGTACTTAATTATGCATTTGAAACGCGCCCAAGTGACGCTAGGCAAGTCAGAGCAGGTTCCCGTGTTAGCTTGAGGGTAAACATACAAGCCGATTGAAGATGGGTAGGGGGCTTCAAATCGTCCAGCACTCCACAGT >candidate_2 AAAACCCATCGTAGTCCTTAGACTTGGGACACTTTCACCCCTAGGGCCCATATCTGGAAATAGACGCCAAGTTCAATCCGTACTCCGACGTACGATGGAACAGTGTGGATGTGACGAGCTTCATTTATACCCTTCGCGCGCCGGACCGGGGTCCGCAAGGCGCGGCGGTG
FASTA headers are optional for a single sequence, but names make your output easier to read.
Sequence length
PETAL models operate on 170-bp windows because the underlying Plant STARR-seq assays used 170-bp sequence fragments.
- Shorter than 170 bp: not enough context for the models. You will see an error!
- Exactly 170 bp: one prediction row.
- Longer than 170 bp: PETAL scans the sequence with overlapping 170-bp windows, moving the window by 1 nt at a time.
How PETAL plots predictions
PETAL models predict the activity of 170-bp sequence windows. How a prediction is displayed therefore depends on the length of the sequence you submitted.
Exactly 170 bp: one prediction
A 170-bp sequence fits the model input exactly, so PETAL makes one prediction per selected condition. PETAL Plot displays these values as a bar plot.
Longer than 170 bp: a sliding-window profile
For a longer sequence, PETAL moves a 170-bp window across the sequence one nucleotide at a time. A sequence of length L therefore produces L - 169 overlapping windows and the same number of predictions per condition. For example, a 200-bp sequence produces 31 prediction windows.
Why does the prediction curve start around position 85?
Each point in a prediction profile represents the activity predicted for an entire 170-bp window, not for a single nucleotide. PETAL places that prediction at the center of its window. Because half of a 170-bp window is 85 bp, a complete window cannot be centered within the first or last 85 bp of the submitted sequence. The prediction profile therefore spans approximately position 85 to position L - 85.
How to read the output
Prediction table
The prediction TSV file contains sequence names, the sequence window, window coordinates and predicted activity columns. Higher values mean higher predicted reporter mRNA levels in the corresponding Plant STARR-seq assay.
Interpret table and plot
Positive attribution scores indicate positions that increase the prediction; negative scores indicate positions that decrease it. Treat strong patterns as candidate motifs or sequence features. They could be protein binding sites.
Evolution table
PETAL Evolve reports the best-performing sequence after each round, its predicted activities and the distance to your requested objective. It can be used to generate candidate sequence variants for later testing.
Things PETAL can and cannot tell you
Good uses
- screen many candidate regulatory fragments
- rank candidate sequences with the same model
- compare conditions from the same model
- generate hypotheses about sequence features
Be careful with
- comparing values across unrelated models
- assuming Interpret peaks are confirmed TF binding sites
- treating evolved sequences as experimentally validated designs
About citations
For publications, please include the PETAL version, the exact model name, the selected conditions and the citation shown after each run. For more background on the training data and model families, see the model details page.
Interested in bringing your model to PETAL?
If you have a model that could be useful to the PETAL community, please reach out to me via email to discuss whether and how your model could be integrated into or hosted on PETAL
PETAL@Home
PETAL@Home is the standalone version of PETAL for running the suite outside this web interface. See the repository for installation instructions, source code and updates.
PETAL@Home repository