What is TRACE?
Why use TRACE?
How do you rank genes per tissue?
What are TRACE values?
Can you support more tissues?
Sometime you refer to a gene and sometime to a protein, do you mean a gene or a protein?
Do you support ranking of gene variants?
What does the example output page show?
How can I contact you?
For additional information:
Contact: Esti Yeger-Lotem estiyl@bgu.ac.il
What is TRACE?
The TRACE webserver takes as input a list of query genes and a tissue, and outputs a TRACE score per gene the reflects its likelihood to lead to a disease that manifests in the selected tissue1. TRACE scores can be used to prioritize the query genes, with higher scores reflecting higher likelihood.
Why use TRACE?
The genetic diagnosis of patients typically yields multiple candidate genes, which complicates the identification of the disease-causing gene. TRACE relies on the observation that many Mendelian and rare diseases manifest clinically in a small subset of tissues2. Thus, it allows users to rank each gene by its likelihood to lead to clinical phenotypes in the disease-affected tissue. By that, TRACE provides a tissue-aware ranking of genes.
TRACE ranking is based on tissue-aware gene features, such as expression per tissue, and not on gene features relating to sequence or conservation, which are used in existing pipelines for analysis and prioritization of candidate disease-causing genes. Thus, TRACE complements existing pipelines and can be used in to refine their output.
How do you rank genes per tissue?
TRACE presents precomputed gene scores per tissue. The computation of scores was achieved via a two-layer procedure that employed multiple machine learning methods and lead to tissue-specific models1. Per model, the TRACE scores of genes were computed using a 10-fold cross validation procedure, as explained in 1.
What are TRACE values?
The TRACE score of a gene reflects the likelihood that an aberration in that gene would lead to a clinical phenotype in the selected tissue. TRACE scores originally ranged between 0 and 1, and were scaled between 0 and 10 to facilitate their interpretation.
Can TRACE support more tissues?
TRACE was applied to tissues for which at least 60 genes that lead to clinical phenotypes per tissue were curated.
Sometime you say gene and sometimes a protein, do you mean a gene or a protein?
Since TRACE relies on features of protein-coding genes that were inferred from datasets of genes or proteins, it associated each protein-coding gene with a single protein product. For simplicity, we refer to the gene and its protein product interchangeably.
Do you support ranking of gene variants?
TRACE does not support ranking of gene variants because it relied on datasets of genes or proteins and not on datasets of variants. Therefore, TRACE treats all variants of the same gene similarly.
What does the example output page show?
The example output page shows the output of TRACE upon uploading a list of 30 genes and selecting the heart tissue model. The output table contains the TRACE scores of the genes in the heart tissue model and their rank relative to each other. The table is ordered such that the top-ranking genes appear first. The top-ranking gene in the heart tissue model is DMD, which is causal for Duchenne muscular dystrophy and dilated cardiomyopathy that manifest in skeletal muscle and heart tissues. The output page allows users to select a different tissue model, which leads to a new output table that contains the TRACE scores and ranks of the same 30 genes in the selected tissue model. For example, upon selecting the brain tissue model, the top-ranking genes are OPA1 that is causal for optic atrophies and SYN1 that is causal for epilepsy with variable learning disabilities and behavior disorders and intellectual developmental disorder.
How can I contact you?
Please email Esti Yeger-Lotem at estiyl@bgu.ac.il
We would love to hear from you!
References
1 Simonovksy, E. et al. Elucidating the tissue-specific impact of hereditary disease genes through interpretable machine learning models. (submitted).
2 Hekselman, I. & Yeger-Lotem, E. Mechanisms of tissue and cell-type specificity in heritable traits and diseases. Nat Rev Genet 21, 137-150, doi:10.1038/s41576-019-0200-9 (2020).