Computational biology and bioinformatics provide vast data gold-mines from protein sequences, ideal for Language Models (LMs) taken from Natural Language Processing (NLP). These LMs reach for new prediction frontiers at low inference costs. Here, we trained two auto-regressive models (Transformer-XL, XLNet) and four auto-encoder models (BERT, Albert, Electra, T5) on data from UniRef and BFD containing up to 393 billion amino acids. The protein LMs (pLMs) were trained on the Summit supercomputer using 5616 GPUs and TPU Pod up-to 1024 cores. Dimensionality reduction revealed that the raw pLM-embeddings from unlabeled data captured some biophysical features of protein sequences. We validated the advantage of using the embeddings as exclusive input for several subsequent tasks: (1) a per-residue (per-token) prediction of protein secondary structure (3-state accuracy Q3=81%-87%); (2) per-protein (pooling) predictions of protein sub-cellular location (ten-state accuracy: Q10=81%) and membrane versus water-soluble (2-state accuracy Q2=91%). For secondary structure, the most informative embeddings (ProtT5) for the first time outperformed the state-of-the-art without multiple sequence alignments (MSAs) or evolutionary information thereby bypassing expensive database searches. Taken together, the results implied that pLMs learned some of the grammar of the language of life. All our models are available through https://github.com/agemagician/ProtTrans.

Inspired by the evolution of biological systems, genetic algorithms have been applied to generate solutions for optimization problems in a variety of scientific and engineering disciplines. For a given problem, a suitable genome representation must be defined along with a mutation operator to generate subsequent generations. Unlike natural systems, which display a variety of complex rearrangements (e.g., mobile genetic elements), mutation for genetic algorithms commonly utilizes only random pointwise changes. Furthermore, generalizing beyond pointwise mutations poses a key difficulty as useful genome rearrangements depend on the representation and problem domain. To move beyond the limitations of manually defined pointwise changes, here we propose the use of techniques from masked language models to automatically generate mutations. As a first step, common subsequences within a given population are used to generate a vocabulary. The vocabulary is then used to tokenize each genome. A masked language model is trained on the tokenized data in order to generate possible rearrangements (i.e., mutations). In order to illustrate the proposed strategy, we use string representations of molecules and use a genetic algorithm to optimize for drug-likeness and synthesizability. Our results show that moving beyond random pointwise mutations accelerates genetic algorithm optimization.

The emergence and rapid worldwide spread of the novel coronavirus disease, COVID-19, has prompted concerted efforts to find successful treatments. The causative virus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), uses its spike (S) protein to gain entry into host cells. Therefore, the S protein presents a viable target to develop a directed therapy. Here, we deployed an integrated artificial intelligence with all-atom molecular dynamics simulation approach to provide new details of the S protein structure. Based on a comprehensive structural analysis of S proteins from SARS-CoV-2 and previous human coronaviruses, we found that the protomer state of S proteins is structurally flexible. Without the presence of a stabilizing beta sheet from another protomer chain, two regions in the S2 domain and the hinge connecting the S1 and S2 subunits lose their secondary structures. Interestingly, the region in the S2 domain was previously identified as an immunodominant site in the SARS-CoV-1 S protein. We anticipate that the molecular details elucidated here will assist in effective therapeutic development for COVID-19.

Molecular Dynamics (MD) simulation have been emerging as an excellent candidate for understanding complex atomic and molecular scale mechanism of bio-molecules that control essential bio-physical phenomenon in a living organism. But this MD technique produces large-size and long-timescale data that are inherently high-dimensional and occupies many terabytes of data. Processing this immense amount of data in a meaningful way is becoming increasingly difficult. Therefore, specific dimensionality reduction algorithm using deep learning technique has been employed here to embed the high-dimensional data in a lower-dimension latent space that still preserves the inherent molecular characteristics i.e. retains biologically meaningful information. Subsequently, the results of the embedding models are visualized for model evaluation and analysis of the extracted underlying features. However, most of the existing visualizations for embeddings have limitations in evaluating the embedding models and understanding the complex simulation data. We propose an interactive visual analytics system for embeddings of MD simulations to not only evaluate and explain an embedding model but also analyze various characteristics of the simulations. Our system enables exploration and discovery of meaningful and semantic embedding results and supports the understanding and evaluation of results by the quantitatively described features of the MD simulations (even without specific labels).

BACKGROUND: We examine the problem of clustering biomolecular simulations using deep learning techniques. Since biomolecular simulation datasets are inherently high dimensional, it is often necessary to build low dimensional representations that can be used to extract quantitative insights into the atomistic mechanisms that underlie complex biological processes. RESULTS: We use a convolutional variational autoencoder (CVAE) to learn low dimensional, biophysically relevant latent features from long time-scale protein folding simulations in an unsupervised manner. We demonstrate our approach on three model protein folding systems, namely Fs-peptide (14 μs aggregate sampling), villin head piece (single trajectory of 125 μs) and β- β- α (BBA) protein (223 + 102 μs sampling across two independent trajectories). In these systems, we show that the CVAE latent features learned correspond to distinct conformational substates along the protein folding pathways. The CVAE model predicts, on average, nearly 89% of all contacts within the folding trajectories correctly, while being able to extract folded, unfolded and potentially misfolded states in an unsupervised manner. Further, the CVAE model can be used to learn latent features of protein folding that can be applied to other independent trajectories, making it particularly attractive for identifying intrinsic features that correspond to conformational substates that share similar structural features. CONCLUSIONS: Together, we show that the CVAE model can quantitatively describe complex biophysical processes such as protein folding.