Protein Complex Structure Prediction Powered by Multiple Sequence Alignments of Interologs from Multiple Taxonomic Ranks and AlphaFold2

Si, Yunda; Yan, Chengfei

doi:10.1101/2021.12.21.473437

Cited by 3 publications

(4 citation statements)

References 33 publications

(51 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As it is trivial to build interologs for them, how to select high-quality MSA for homodimers is a more challenging yet important question. Previous work [39, 54] has an empirical insight that instead of using the full MSA searched from the protein sequence database, we can select a few high-quality MSA following some promisings, such as using the MSA maximizing the sequence diversity [39], or choosing the MSA owning the largest sequence similarity with the primary sequence [54]. To date, few efforts have systematically investigated the MSA-selection problem.…”

Section: Conclusion and Discussionmentioning

confidence: 99%

“…KM algorithm finds a global optimal solution. However, as suggested by [54], in each species, the sequence that is most similar to the query sequence may be more informative, while other sequences that are less similar may add noises. Thus we propose a greedy algorithm that focuses on pairs that have high similarity scores with the query sequence.…”

Section: Local Maximization Optimization (Interlocalcos)mentioning

confidence: 99%

See 1 more Smart Citation

Improved the Protein Complex Prediction with Protein Language Models

Chen

Xie

et al. 2022

Preprint

View full text Add to dashboard Cite

AlphaFold-Multimer has greatly improved protein complex structure prediction, but its accuracy also depends on the quality of the multiple sequence alignment (MSA) formed by the interacting homologs (i.e., interologs) of the complex under prediction. Here we propose a novel method, denoted as ColAttn, that can identify interologs of a complex by making use of protein language models (PLMs). We show that ColAttn can generate better interologs than the default MSA generation method in AlphaFold-Multimer. Our method results in better complex structure prediction than AlphaFold-Multimer by a large margin (+10.7% in terms of the Top-5 best DockQ), especially when the predicted complex structures have low confidence. We further show that by combining several MSA generation methods, we may yield even better complex structure prediction accuracy than Alphafold-Multimer (+22% in terms of the Top-5 best DockQ). We systematically analyze the impact factors of our algorithm and find out the diversity of MSA of interologs significantly affects the prediction accuracy. Moreover, we show that ColAttn performs particularly well on complexes in eukaryotes.

show abstract

Section: Conclusion and Discussionmentioning

confidence: 99%

Section: Local Maximization Optimization (Interlocalcos)mentioning

confidence: 99%

Improved the Protein Complex Prediction with Protein Language Models

Chen

Xie

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…For homomeric PPIs, the paired MSA is formed by concatenating two copies of the MSA. For heteromeric PPIs, the paired MSA is formed by pairing the MSAs through the phylogeny-based approach described in (https://github.com/ChengfeiYan/PPI_MSA-taxonomy_rank) 46 . We input the paired MSA into CCMpred 47 to get the evolutionary coupling matrix, and into alnstats 48 to get mutual information matrix, APC-corrected mutual information matrix and contact potential matrix.…”

Section: Methodsmentioning

confidence: 99%

Protein language model embedded geometric graphs power inter-protein contact prediction

Yan

2023

Preprint

Self Cite

View full text Add to dashboard Cite

Accurate prediction of contacting residue pairs between interacting proteins is very useful for structural characterization of protein-protein interactions (PPIs). Although significant improvement has been made in inter-protein contact prediction recently, there is still large room for improving the prediction accuracy. Here we present a new deep learning method referred to as PLMGraph-Inter for inter-protein contact prediction. Specifically, we employ rotationally and translationally invariant geometric graphs obtained from structures of interacting proteins to integrate multiple protein language models, which are successively transformed by graph encoders formed by geometric vector perceptrons and residual networks formed by dimensional hybrid residual blocks to predict inter-protein contacts. Extensive evaluation on multiple test sets shows that PLMGraph-Inter significantly outperforms three top inter-protein contact prediction methods, including DRN-1D2D_Inter, DeepHomo and GLINTER. Finally, we show leveraging the contacts predicted by PLMGraph-Inter as constraints for protein-protein docking can dramatically improve its performance for protein complex structure prediction.

show abstract

“…The paired MSA for each homomeric PPI was formed by concatenating two copies of the MSA. For heteromeric PPIs, the paired MSA was formed by pairing the MSAs through the phylogeny-based approach described in 27 .…”

Section: The Preparation Of Input Featuresmentioning

confidence: 99%

Improved inter-protein contact prediction using dimensional hybrid residual networks and protein language models

Yan

2022

Preprint

Self Cite

View full text Add to dashboard Cite

The knowledge of contacting residue pairs between interacting proteins is very useful for structural characterization of protein-protein interactions (PPIs). However, accurately identifying the tens of contacting ones from hundreds of thousands of inter-protein residue pairs is extremely challenging, and performances of the state-of-the-art inter-protein contact prediction methods are still quite limited. In this study, we developed a deep learning method for inter-protein contact prediction, referred to as DRN-1D2D_Inter. Specifically, we employed pretrained protein language models to generate structural information enriched input features to residual networks formed by dimensional hybrid residual blocks to perform inter-protein contact prediction. Extensively benchmarked DRN-1D2D_Inter on multiple datasets including both heteromeric PPIs and homomeric PPIs, we show DRN-1D2D_Inter consistently and significantly outperformed two state-of-the-art inter-protein contact prediction methods including GLINTER and DeepHomo, although both the latter two methods leveraged native structures of interacting proteins in the contact prediction, and DRN-1D2D_Inter made the prediction purely from sequences.

show abstract

Protein Complex Structure Prediction Powered by Multiple Sequence Alignments of Interologs from Multiple Taxonomic Ranks and AlphaFold2

Cited by 3 publications

References 33 publications

Improved the Protein Complex Prediction with Protein Language Models

Improved the Protein Complex Prediction with Protein Language Models

Protein language model embedded geometric graphs power inter-protein contact prediction

Improved inter-protein contact prediction using dimensional hybrid residual networks and protein language models

Contact Info

Product

Resources

About