Welcome to BRASS

BRcA Specific Software for predicting the pathogenicity of BRCA1 and BRCA2 variants

In silico predictions of BRCA1 and BRCA2 variants

BRASS offers several predictors that reflect either the functional or clinical impact of a given variant


BRCA Specific Software (BRASS) for predicting the impact of nonsynonymous variants in BRCA1


BRCA Specific Software (BRASS) for predicting the impact of nonsynonymous variants in BRCA2

Disclaimer This resource is uniquely intended for research purposes. The authors are not responsible for neither its use nor misuse. The data provided are not intended as advice of any kind. The authors have worked with care in the development of this server, but assume no liability or responsibility for any error, weakness, incompleteness or temporariness of the resource and of the data provided.

About BRASS predictions and output

For a given variant, BRASS provides a pathogenicity prediction obtained either from our Neural Network (NN) predictor or from our Multiple Linear Regression (MLR) predictor of the functional impact of the HDR assay. The output is similar in both cases, and is described below.

Label Plot

Binary output: the pathogenicity prediction

Protein sequence variants are classified as pathogenic or neutral by the predictor.

This output is a discretization of the numerical score provided by the predictor, using a pre-established threshold. This threshold is 0.5 for the NN predictor, 0.53 for the HDR-based BRCA1 predictor, and 2.25 for the HDR-based BRCA2 predictor.

Score Plot

Beyond the binary output: the numerical score

This score will provide you with a quantitative view of the prediction

For the NN predictor, the score is a continous value comprised between 0 (neutral) and 1 (pathogenic). For the HDR-based predictors, the output is an estimate of the result of the HDR experiment for the corresponding protein. It is therefore a directly interpretable quantity, as the actual experiment would be. Predictions below the threshold are pathogenic whereas predictions above the threshold are neutral.

Reliability Plot

Prediction Reliability

This metric gives an idea of the prediction accuracy

For the NN predictor we provide a simple relability measure, which is encoded using 5 circles. The more filled circles, the more reliable is the prediction.

Performance Plot

Predictor Performance

Estimates of the predictors' performances can be found here for BRCA1 and here for BRCA2.

They have been obtained following a standard leave-one-out cross-validation procedure.

Supporting Institutions

Vall d'Hebron
Asociación Española Contra el Cancer
Instituto de Salud Carlos III
Ministerio de Economía y Empresa