The BUSCA web server is described in:

Savojardo, C, Martelli, PL, Fariselli, P, Profiti, G, Casadio, R

Nucleid Acids Research (2018), 46(W1), W459-W466.

Predictors included into BUSCA

DeepSig - Prediction of secretory signal peptides.

Savojardo, C., Martelli P.L., Fariselli, P., Casadio, R. (2018) DeepSig: deep learning improves signal peptide detection. Bioinformatics, 34(10), 1690-1696.

A new method that takes advantage of Deep Learning and improves the state-of-the-art performance. DeepSig is designed for both detecting signal peptides and finding their cleavage sites in protein sequences. The predictor consists of two consecutive building blocks: a deep neural network architecture and a probabilistic method that incorporates the current biological knowledge of the signal peptide structure. DeepSig is optimized to discriminate true signal peptide sequences from similar N-terminal transmembrane regions. This is accomplished by devising a Deep Convolutional Neural Network comprising three cascading convolution-pooling stages that process the N-terminus of the query protein, sorting out 3 classes: signal peptides, transmembrane regions, and anything else.

SChloro - Prediction of sub-chloroplastic localization of proteins.

Savojardo, C., Martelli, P.L., Fariselli, P., Casadio, R. (2017) SChloro: directing Viridiplantae proteins to six chloroplastic sub-compartments, Bioinformatics, 33(3), 347-353.

Predicts protein sub-chloroplastic localization, based on targeting signal detection and membrane protein information. SChloro is based on the recognition of sequence signals that define target specificity (chloroplast and thylakoid targeting signals) as well as on the prediction of the potential type of interaction with chloroplast membranes (single-pass, multi-pass and peripheral interaction). All these features are processed using a multi-stage architecture based on Support Vector Machines. SChloro can also predict multiple sub-chloroplast localizations for the same protein.

TPpred3 - Prediction of organelle-targeting peptides.

Savojardo, C., Martelli, P.L., Fariselli, P., Casadio, R. (2015) TPpred3 detects and discriminates mitochondrial and chloroplastic targeting peptides in eukaryotic proteins, Bioinformatics, 31(20), 3269-3275.

A pipeline for predicting organelle-targeting peptides based on different machine learning approaches (Conditional Random Fields, Neural Networks and Support Vector Machines). TPpred3 is a multi-stage method which detects targeting peptides at the N-terminus of the query protein sequence, discriminates chroloplastic from mitochondrial peptides (in plants) and ultimately identifies the peptide cleavage site, exploiting organelle-specific short sequence motifs sorrounding the site.

BetAware - Recognition and topology prediction of prokaryotic beta-barrel membrane proteins.

Savojardo, C., Fariselli, P., Casadio, R. (2013) BETAWARE: a machine-learning tool to detect and predict transmembrane beta-barrel proteins in prokaryotes. Bioinformatics, 29(4), 504-505.

BetAware is a method for TransMembrane β-Barrel (TMBB) detection and topology prediction. Both prediction steps are based on advanced machine-learning methods. For TMBB detection, BetAware exploits a new machine learning approach based on N-to-1 Extreme Leaning Machines, while TMBB topology prediction is carried-out using a probabilistic model based on Grammatical-Restrained Hidden Conditional Random Fields.

MemLoci - Prediction of membrane protein localization.

Pierleoni, A., Martelli, P.L., Casadio, R. (2011) MemLoci: predicting subcellular localization of membrane proteins in eukaryotes. Bioinformatics, 27(9), 1224-1230.

MemLoci is a Support Vector Machine based predictor specifically trained to discriminate the subcellular localization of proteins associated or inserted in eukaryotes membranes. It is able to predict whether a protein sequence is localized to the plasma membrane, to organelle membranes or to other intracellular organelle membranes.

PredGPI - Prediction of GPI-anchors in proteins.

Pierleoni, A., Martelli, P.L., Casadio, R. (2008) PredGPI: a GPI-anchor predictor. BMC Bioinformatics, 9, 392.

PredGPI is a prediction system for GPI-anchored proteins. It is based on a support vector machine (SVM) for the discrimination of the anchoring signal, and on a Hidden Markov Model (HMM) for the prediction of the most probable omega-site.

BaCelLo - Prediction of protein subcellular localizations.

Pierleoni, A., Martelli, P.L., Fariselli, P., Casadio, R. (2006) BaCelLo: a balanced subcellular localization predictor, Bioinformatics, 22(14), e408-e416.

BaCelLo is a predictor for the subcellular localization of proteins in eukaryotes. It is based on a decision tree of several support vector machines (SVMs), it classifies up to four localizations for Fungi and Metazoan proteins and five localizations for Plant ones. BaCelLo's predictions are balanced among different classes and all the localizations are considered as equiprobable.

ENSEMBLE3.0 - Recognition and topology prediction of all-alpha membrane proteins.

Martelli, P.L., Fariselli, P., Casadio, R. (2003) An ENSEMBLE machine learning approach for the prediction of all-alpha membrane proteins. Bioinformatics, 19(Suppl 1), 205-211.

The last version of the ENSEMBLE tool for predicting the topology of all-alpha membrane proteins (Martelli et al., 2003). In its present version, ENSEMBLE3.0 predicts the location of transmembrane alpha-helices along the sequence as well as the classification between transmembrane and globular proteins.