TraitConnect · PearGRN

TraitConnect · trait probabilities from the network

Ask which genes look like they belong to a trait. Every gene in the pear genome has already been scored by five machine-learning models trained on its network neighbourhood combined with expression and protein data, so this page returns pre-computed probabilities straight away. Paste the gene IDs you care about, or rank the whole genome at once.

Genes scored

17,868

Prediction sets

62 traits × 3 omics

Models per gene

5plus their average

Trait what the model predicts

Sugar metabolism Genes whose network neighbourhood predicts a role in sugar metabolism. Lignin metabolism Genes whose network neighbourhood predicts a role in lignin metabolism.

Omics level which measurements the model was trained on

Network + all omics network topology together with gene expression and protein abundance Network + gene expression network topology together with gene expression only Network + protein abundance network topology together with protein abundance only

Gene IDs one per line, duplicates are ignored

Examples

How to read the result

Each row is one gene. The five model columns hold that model's probability that the gene takes part in the selected trait, on a 0–1 scale — higher means more likely. Average is the mean of the five and is what the table is sorted by, so the strongest candidates sit on top.

The models disagree by design: a gene that scores high under every model is a safer candidate than one carried by a single column.

All Genes ranks every gene in the genome and is capped at 2,000 rows — open the full list from the result page when you need it.

Six prediction sets exist in total: each trait is scored against network topology combined with all omics, with expression only, and with protein abundance only. Switching the omics level changes the table, so the same gene can score differently.