Being able to predict the course of arbitrary chemical reactions is essential to the theory and applications of organic chemistry. Previous approaches are not high-throughput, are not generalizable or scalable, or lack sufficient data to be effective. We describe single mechanistic reactions as concerted electron movements from an electron orbital source to an electron orbital sink. We use an existing rule-based expert system to derive a dataset consisting of 2,989 productive mechanistic steps and 6.14 million non-productive mechanistic steps. We then pose identifying productive mechanistic steps as a ranking problem: rank potential orbital interactions such that the top ranked interactions yield the major products. The machine learning implementation follows a two-stage approach, in which we first train atom level reactivity filters to prune 94.0% of non-productive reactions with less than a 0.1% false negative rate. Then, we train an ensemble of ranking models on pairs of interacting orbitals to learn a relative productivity function over single mechanistic reactions in a given system. Without the use of explicit transformation patterns, the ensemble perfectly ranks the productive mechanisms at the top 89.1% of the time, rising to 99.9% of the time when top ranked lists with at most four non-productive reactions are considered. The final system allows multi-step reaction prediction. Furthermore, it is generalizable, making reasonable predictions over reactants and conditions which the rule-based expert system does not handle.

MOTIVATION: Discovery of novel protective antigens is fundamental to the development of vaccines for existing and emerging pathogens. Most computational methods for predicting protein antigenicity rely directly on homology with previously characterized protective antigens; however, homology-based methods will fail to discover truly novel protective antigens. Thus, there is a significant need for homology-free methods capable of screening entire proteomes for the antigens most likely to generate a protective humoral immune response. RESULTS: Here we begin by curating two types of positive data: (i) antigens that elicit a strong antibody response in protected individuals but not in unprotected individuals, using human immunoglobulin reactivity data obtained from protein microarray analyses; and (ii) known protective antigens from the literature. The resulting datasets are used to train a sequence-based prediction model, ANTIGENpro, to predict the likelihood that a protein is a protective antigen. ANTIGENpro correctly classifies 82% of the known protective antigens when trained using only the protein microarray datasets. The accuracy on the combined dataset is estimated at 76% by cross-validation experiments. Finally, ANTIGENpro performs well when evaluated on an external pathogen proteome for which protein microarray data were obtained after the initial development of ANTIGENpro. AVAILABILITY: ANTIGENpro is integrated in the SCRATCH suite of predictors available at http://scratch.proteomics.ics.uci.edu. CONTACT: [email protected]