Sign in to PlantQTLdb Assistant
Connect molecular QTL, GWAS and literature evidence in a source-linked explanation.
- Which evidence supports a gene–trait association?
- How do molecular QTL and literature findings compare?
A PlantQTLdb account is required to start a conversation.
Welcome to PlantQTLdb Assistant
Gene-centered interpretation of molecular QTL, GWAS, and literature evidence across plant species
Developer run details
No execution yet.
Sequence scores
| Record | Bases | GC % | NLL ↓ | Perplexity ↓ |
|---|
Sequence logo and observed bases
Each letter’s height = its DNA-normalized probability × stack information. The row below is your observed sequence. A tall letter is a model preference, not an experimentally supported annotation.
The first part is your original prompt; the following 200 bases are model-generated. This task does not score or annotate the prompt. A generated continuation is not a reconstruction of the true reference genome.
What do these results mean, and when should I use Evo2?
Scoring: use a genomic fragment to explore model predictability. Lower NLL means the actual input is more expected; perplexity = exp(NLL). A score of 0.87 is not 87% accuracy or a probability of biological function.
Entropy and logo: higher entropy means more uncertainty across possible next tokens. The logo shows preferences within A/C/G/T; taller stacks indicate more concentrated preferences, not measured evolutionary conservation. A confident model can still assign low probability to your actual base.
Comparison: submit equal-length reference and alternative FASTA records with matching flanks. Positive ΔNLL means the alternative is less expected; this can inform follow-up priorities, but does not establish harm or a splice effect.
Completion: explicitly request generation with an Arabidopsis or maize condition and a 2–500 bp prompt. The 200 generated bases are candidates, not a reconstruction of missing genomic sequence.
These scores and plots do not label exons or introns. Use reference annotation and transcript evidence for gene structure.
Official coding-region annotation and protein structure workflow: separate annotation methods; ESMFold visualization is described for prokaryotic sequences. This tool does not collect those outputs.
Official mechanistic visualizer: explore existing bacterial genome examples; this is not custom plant-sequence analysis.