Protein language models can capture protein quaternary state

Abstract

Background Determining a protein’s quaternary state, i.e. the number of monomers in a functional unit, is a critical step in protein characterization. Many proteins form multimers for their activity, and over 50% are estimated to naturally form homomultimers. Experimental quaternary state determination can be challenging and require extensive work. To complement these efforts, a number of computational tools have been developed for quaternary state prediction, often utilizing experimentally validated structural information. Recently, dramatic advances have been made in the field of deep learning for predicting protein structure and other characteristics. Protein language models, such as ESM-2, that apply computational natural-language models to proteins successfully capture secondary structure, protein cell localization and other characteristics, from a single sequence. Here we hypothesize that information about the protein quaternary state may be contained within protein sequences as well, allowing us to benefit from these novel approaches in the context of quaternary state prediction. Results We generated ESM-2 embeddings for a large dataset of proteins with quaternary state labels from the curated QSbio dataset. We trained a model for quaternary state classification and assessed it on a non-overlapping set of distinct folds (ECOD family level). Our model, named QUEEN (QUaternary state prediction using dEEp learNing), performs worse than approaches that include information from solved crystal structures. However, it successfully learned to distinguish multimers from monomers, and predicts the specific quaternary state with moderate success, better than simple sequence similarity-based annotation transfer. Our results demonstrate that complex, quaternary state related information is included in such embeddings. Conclusions QUEEN is the first to investigate the power of embeddings for the prediction of the quaternary state of proteins. As such, it lays out strengths as well as limitations of a sequence-based protein language model approach, compared to structure-based approaches. Since it does not require any structural information and is fast, we anticipate that it will be of wide use both for in-depth investigation of specific systems, as well as for studies of large sets of protein sequences. A simple colab implementation is available at: https://colab.research.google.com/github/Furman-Lab/QUEEN/blob/main/QUEEN_prediction_notebook.ipynb.

Transformer-Based Deep Learning Model with Latent Space Regularization for CRISPR-Cas Protein Sequence Classification

“…Furthermore, transformer models have revolutionized the field by enabling the development of large Protein Language Models (LPLMs), which have emerged as transformative tools in computational biology and bioinformatics [23,24]. Similar to large Protein Language Models (LLMs) of NLP trained on large corpora of words [25], similar efforts have been applied to train LPLMs using large protein databases such as BFD100, UniRef50, and UniRef100 with trillions of protein sequences.…”

DeepSub: Utilizing Deep Learning for Predicting the Number of Subunits in Homo-Oligomeric Protein Complexes

“…At the same time, there is also a considerable proportion of proteins with over 30% similarity but different NS. So, relying solely on sequence similarity (e.g., a threshold of 30%) to predict the oligomeric state of proteins may not always be reliable, a point that has been mentioned in earlier research [21] as well.…”

“…Recent advancements in deep learning have shown promise in predicting protein's quaternary state. Protein language models, utilizing computational natural language processing techniques for proteins, have successfully captured secondary structure, protein cellular localization, and other features from amino acid sequences [21]. This raises the question: can a protein's quaternary state be inferred solely from its sequence?…”

“…This raises the question: can a protein's quaternary state be inferred solely from its sequence? Orly et al introduced "QUEEN" [21], a study exploring the use of the pretrained model Evolutionary Scale Modeling 2 (ESM2) [22] for predicting protein quaternary state from sequences. However, we have observed over 50% of protein fragments in the training dataset.…”

Rapid and accurate prediction of protein homo-oligomer symmetry with Seq2Symm

“…More computationally efficient methods to fold large protein oligomers, such as Uni-Fold Symmetry [ 13 ] still require the pre-specified symmetry group as input to make predictions. Protein embeddings from ESM2 [ 14 ] have been used to predict the most likely quaternary state of a protein chain (QUEEN [ 15 ]); however, in this approach, the model only predicts the multiplicity of the oligomer thereby giving no clue as to global symmetry of the protein.…”