rawMSA: End-to-end Deep Learning Makes Protein Sequence Profiles and Feature Extraction obsolete

In the last few decades, huge efforts have been made in the bioinformatics community to develop machine learning-based methods for the prediction of structural features of proteins in the hope of answering fundamental questions about the way proteins function and about their involvement in several illnesses. The recent advent of Deep Learning has renewed the interest in neural networks, with dozens of methods being developed in the hope of taking advantage of these new architectures. On the other hand, most methods are still based on heavy pre-processing of the input data, as well as the extraction and integration of multiple hand-picked, manually designed features. Since Multiple Sequence Alignments (MSA) are almost always the main source of information in de novo prediction methods, it should be possible to develop Deep Networks to automatically refine the data and extract useful features from it. In this work, we propose a new paradigm for the prediction of protein structural features called rawMSA. The core idea behind rawMSA is borrowed from the field of natural language processing to map amino acid sequences into an adaptively learned continuous space. This allows the whole MSA to be input into a Deep Network, thus rendering sequence profiles and other pre-calculated features obsolete. We developed rawMSA in three different flavors to predict secondary structure, relative solvent accessibility and inter-residue contact maps. We have rigorously trained and benchmarked rawMSA on a large set of proteins and have determined that it outperforms classical methods based on position-specific scoring matrices (PSSM) when predicting secondary structure and solvent accessibility, while performing on a par with the top ranked CASP12 methods in the inter-residue contact map prediction category. We believe that rawMSA represents a promising, more powerful approach to protein structure prediction that could replace older methods based on protein profiles in the coming years. Availability datasets, dataset generation code, evaluation code and models are available at: https://bitbucket.org/clami66/rawmsa