Recently, we developed SPIDER3-Single that was dedicated to single-sequence-based prediction for not only secondary structure and solvent accessibility but also other structural properties such as backbone torsion angles and half-sphere exposures (HSE) ( Heffernan et al., 2018). A method called accessible surface area (ASA)-quick employed single-sequence information to predict the Accessible Surface Area ( Faraggi et al., 2017). (2000) proposed a profile-based secondary structure predictor along with a single-sequence-based predictor, which we will refer to as PSIPRED-Single. Unlike evolution profile-based methods, only a few methods were developed for single-sequence-based prediction of one-dimensional structural properties. After all, proteins fold into secondary and tertiary structures from their single sequences only. Developing single-sequence-based methods is important because one can only claim that the problem of secondary structure prediction is solved if only a single sequence is utilized as an input for prediction. Thus, it is necessary to develop single-sequence or no evolutionary information-based methods. Their predicted secondary structure and other structural properties are substantially less accurate than those with many homologous sequences ( Heffernan et al., 2018). However, more than 90% of proteins have none or very few homologous sequences ( Ovchinnikov et al., 2017). These methods employed the feature profiles generated by PSI-BLAST ( Altschul et al., 1997), HHblits ( Remmert et al., 2012) and the output of other predictors ( Cuff and Barton, 2000). However, most of the above-stated improvement came from evolutionary-profile-based methods ( Hanson et al., 2019 Klausen et al., 2019 Wang et al., 2016 Xu et al., 2020). Particularly, the protein secondary structure prediction is approaching the theoretical upper bounds at 88–90% accuracy with SPOT-1D prediction of three-state secondary structure at 86.18% and eight-state secondary structure at 79% accuracy ( Hanson et al., 2019 Yang et al., 2018). These improvements have ultimately led to a considerable improvement in protein tertiary structure prediction, as observed in CASP13 ( Cheng et al., 2019). Significant headway has been observed specifically for the protein secondary structure and contact map prediction ( Fang et al., 2018 Hanson et al., 2019 Li et al., 2019 Wang et al., 2016 Wu et al., 2020). The past two decades have seen many developments in the field of deep learning-based prediction of protein structure ( Yang et al., 2018).
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |