Biology subjects

Onoda, A.

Publications and source records attributed to Onoda, A..

3 recordsLinked to original sources

bioRxiv2024

Accurate and Rapid Prediction of Protein pKa: Protein Language Models Reveal the Sequence-pKa Relationship

Protein pKa prediction is a key challenge in computational biology. In this study, we present pKALM, a novel deep learning-based method for high-throughput protein pKa prediction. pKALM uses a protein language model (PLM) to capture the complex sequence-structure relationship of proteins. While traditionally considered a structure-based problem, our results show that a PLM pre-trained on large-scale protein sequence databases can effectively learn this relationship and achieve state-of-the-art performance. pKALM accurately predicts the pKa values of six residues (Asp, Glu, His, Lys, Cys, and Tyr) and two termini with high precision and efficiency. It excels at predicting both exposed and buried residues, which often deviate from standard pKa values measured in solvent. We demonstrate a novel finding that predicted protein isoelectric points (pI) can be used to improve the accuracy of pKa prediction. High-throughput pKa prediction of the human proteome using pKALM achieves a speed of 4,965 pKa predictions per second, which is several orders of magnitude faster than existing state-of-the-art methods. The case studies illustrate the efficacy of pKALM in estimating pKa values and the constraints of the method. pKALM will thus be a valuable tool for researchers in the fields of biochemistry, biophysics, and drug design.

bioinformatics↗

bioRxiv2024

PsiPartition: Improved Site Partitioning for Genomic Data by Parameterized Sorting Indices and Bayesian Optimization

Phylogenetics has been widely used in molecular biology to infer the evolutionary relationships among species. With the rapid development of sequencing technology, genomic data with thousands of sites becomes increasingly common in phylogenetic analysis, while heterogeneity among sites arises as one of the major challenges. A single homogeneous model is not sufficient to describe the evolution of all sites and partitioned models are often employed to model the evolution of heterogeneous sites by partitioning them into distinct groups and utilizing distinct evolutionary models for each group. It is crucial to determine the best partitioning, which greatly affects the reconstruction correctness of phylogeny. However, the best partitioning is usually intractable to obtain in practice. Traditional partitioning methods rely on heuristic algorithms or greedy search to determine the best ones in their solution space, are usually time-consuming, and with no guarantee of optimality. In this study, we propose a novel partitioning approach, termed PsiPartition, based on the parameterized sorting indices of sites and Bayesian optimization. We apply our method to empirical data sets and it performs significantly better compared to existing methods, in terms of Bayesian information criterion (BIC) and the corrected Akaike information criterion (AICc). We test PsiPartition on the simulated data sets with different site heterogeneity, alignment lengths, and number of loci. It is demonstrated that PsiPartition evidently and stably outperforms other methods in terms of the Robinson-Foulds (RF) distance between the true simulated trees and the reconstructed trees, especially on the data with more site heterogeneity. More importantly, our proposed Bayesian optimization-based method, for the first time, provides a new general framework to efficiently determine the optimal number of partitions. The corresponding reproducible source code and data are available at http://github.com/xu-shi-jie/PsiPartition.

bioinformatics↗

bioRxiv2022

Fast and Accurate Prediction of Intrinsically Disordered Protein by Protein Language Model

MotivationIntrinsically disordered proteins (IDPs) play a vital role in various biological processes and have attracted increasing attention in the last decades. Predicting IDPs from primary structures of proteins provides a very useful tool for protein analysis. However, most of the existing prediction methods heavily rely on multiple sequence alignments (MSAs) of homologous sequences which are formed by evolution over billions of years. Obtaining such information requires searching against the whole protein databases to find similar sequences and since this process becomes increasingly time-consuming, especially in large-scale practical applications, the alternative method is needed. ResultsIn this paper, we proposed a novel IDP prediction method named IDP-PLM, based on the protein language model (PLM). The method does not rely on MSAs or MSA-based profiles but leverages only the protein sequences, thereby achieving state-of-the-art performance even compared with predictors using protein profiles. The proposed IDP-PLM is composed of stacked predictors designed for several different protein-related tasks: secondary structure prediction, linker prediction, and binding predictions. In addition, predictors for the single task also achieved the highest accuracy. All these are based on PLMs thus making IDP-PLM not rely on MSA-based profiles. The ablation study reveals that all these stacked predictors contribute positively to the IDP prediction performance of IDP-PLM. AvailabilityThe method is available at http://github.com/xu-shi-jie. Contactakira.onoda@ees.hokudai.ac.jp Supplementary informationSupplementary data are available at Bioinformatics online.

bioinformatics↗