Bagging MSA Learning: Enhancing Low-Quality PSSM with Deep Learning for Accurate Protein Structure Property Prediction

Guo, Yuzhi; Wu, Jiaxiang; Ma, Hehuan; Wang, Sheng; Huang, Junzhou

doi:10.1007/978-3-030-45257-5_6

Cited by 17 publications

(15 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each amino acid in the protein sequence is represented by a one-hot vector with length as 21, which refers to 20 kinds of amino acids plus one unknown amino acid. PSSM represents the distribution of amino acid types on each position in the protein sequence [ 46 ]. Following the same procedure in [ 5 , 41 , 47 ], we get the PSSM matrix by searching Uniref50 database [ 48 ], and concatenate it with the one-hot vectors.…”

Section: Methodsmentioning

confidence: 99%

Deep Ensemble Learning with Atrous Spatial Pyramid Networks for Protein Secondary Structure Prediction

Guo

Wu²,

Ma³

et al. 2022

Biomolecules

Self Cite

View full text Add to dashboard Cite

The secondary structure of proteins is significant for studying the three-dimensional structure and functions of proteins. Several models from image understanding and natural language modeling have been successfully adapted in the protein sequence study area, such as Long Short-term Memory (LSTM) network and Convolutional Neural Network (CNN). Recently, Gated Convolutional Neural Network (GCNN) has been proposed for natural language processing. It has achieved high levels of sentence scoring, as well as reduced the latency. Conditionally Parameterized Convolution (CondConv) is another novel study which has gained great success in the image processing area. Compared with vanilla CNN, CondConv uses extra sample-dependant modules to conditionally adjust the convolutional network. In this paper, we propose a novel Conditionally Parameterized Convolutional network (CondGCNN) which utilizes the power of both CondConv and GCNN. CondGCNN leverages an ensemble encoder to combine the capabilities of both LSTM and CondGCNN to encode protein sequences by better capturing protein sequential features. In addition, we explore the similarity between the secondary structure prediction problem and the image segmentation problem, and propose an ASP network (Atrous Spatial Pyramid Pooling (ASPP) based network) to capture fine boundary details in secondary structure. Extensive experiments show that the proposed method can achieve higher performance on protein secondary structure prediction task than existing methods on CB513, Casp11, CASP12, CASP13, and CASP14 datasets. We also conducted ablation studies over each component to verify the effectiveness. Our method is expected to be useful for any protein related prediction tasks, which is not limited to protein secondary structure prediction.

show abstract

Section: Methodsmentioning

confidence: 99%

Deep Ensemble Learning with Atrous Spatial Pyramid Networks for Protein Secondary Structure Prediction

Guo

Wu²,

Ma³

et al. 2022

Biomolecules

Self Cite

View full text Add to dashboard Cite

show abstract

“…The associative online material is available at: https://github.com/xiaozhi0689/EPTool, which describes the detailed tutorial for using EPTool. The functionality and specific instructions, as well as examples for (5) utilizing other data sets to train the enhancing PSSM model from scratch, details about the model implementation may refer to Bagging MSA (Guo et al, 2020).…”

Section: Methodsmentioning

confidence: 99%

EPTool: A New Enhancing PSSM Tool for Protein Secondary Structure Prediction

Guo

et al. 2021

Journal of Computational Biology

Self Cite

View full text Add to dashboard Cite

Recently, a deep learning-based enhancing Position-Specific Scoring Matrix (PSSM) method (Bagging Multiple Sequence Alignment [MSA] Learning) Guo et al. has been proposed, and its effectiveness has been empirically proved. Program EPTool is the implementation of Bagging MSA Learning, which provides a complete training and evaluation workflow for the enhancing PSSM model. It is capable of handling different input data set and various computing algorithms to train the enhancing model, then eventually improve the PSSM quality for those proteins with insufficient homologous sequences. In addition, EPTool equips several convenient applications, such as PSSM features calculator, and PSSM features visualization. In this article, we propose designed EPTool and briefly introduce its functionalities and applications. The detailed accessible instructions are also provided.

show abstract

“…However, those approaches still require rich MSA features to provide co-evolution knowledge. (Guo and et al 2020) firstly proposes a selfsupervised approach to predict an enhanced PSSM from low quality PSSM to improve the protein secondary structure prediction. (Wang and et al 2021b) further introduces knowledge distillation and contrastive learning to jointly optimize the enhanced network and secondary structure predictor.…”

Section: Low Homologous Protein Structure Predictionmentioning

confidence: 99%

Contact-Distil: Boosting Low Homologous Protein Contact Map Prediction by Self-Supervised Distillation

Wang

Chen

Zhou

et al. 2022

AAAI

Self Cite

View full text Add to dashboard Cite

Accurate protein contact map prediction (PCMP) is essential for precise protein structure estimation and further biological studies. Recent works achieve significant performance on this task with high quality multiple sequence alignment (MSA). However, the PCMP accuracy drops dramatically while only poor MSA (e.g., absolute MSA count less than 10) is available. Therefore, in this paper, we propose the Contact-Distil to improve the low homologous PCMP accuracy through knowledge distillation on a self-supervised model. Particularly, two pre-trained transformers are exploited to learn the high quality and low quality MSA representation in parallel for the teacher and student model correspondingly. Besides, the co-evolution information is further extracted from pure sequence through a pretrained ESM-1b model, which provides auxiliary knowledge to improve student performance. Extensive experiments show Contact-Distil outperforms previous state-of-the-arts by large margins on CAMEO-L dataset for low homologous PCMP, i.e., around 13.3% and 9.5% improvements against Alphafold2 and MSA Transformer respectively when MSA count less than 10.

show abstract

Bagging MSA Learning: Enhancing Low-Quality PSSM with Deep Learning for Accurate Protein Structure Property Prediction

Cited by 17 publications

References 32 publications

Deep Ensemble Learning with Atrous Spatial Pyramid Networks for Protein Secondary Structure Prediction

Deep Ensemble Learning with Atrous Spatial Pyramid Networks for Protein Secondary Structure Prediction

EPTool: A New Enhancing PSSM Tool for Protein Secondary Structure Prediction

Contact-Distil: Boosting Low Homologous Protein Contact Map Prediction by Self-Supervised Distillation

Contact Info

Product

Resources

About