# Jpred Jpred is a protein secondary structure prediction server developed at the University of Dundee, providing computational predictions for protein sequences via a web interface. It uses a combination of machine learning and evolutionary information from multiple sequence alignments. Jpred is a web-based server for protein secondary structure prediction, developed and maintained by the Barton Group at the University of Dundee. It accepts a protein sequence and returns a prediction of the secondary structure (alpha-helix, beta-strand, or coil) along with solvent accessibility and confidence scores. The server has been publicly available since the late 1990s and is widely used in bioinformatics research. The prediction method in Jpred integrates multiple sequence alignments generated by PSI-BLAST against protein databases, capturing evolutionary information that improves accuracy. The core algorithm employs a neural network, specifically a recurrent architecture, trained on known protein structures. Over the years, Jpred has undergone several updates, with the current version (Jpred 4) offering improved performance and a user-friendly interface. ## History and Development Jpred was first released in 1998 by James Cuff and Geoffrey Barton at the University of Dundee. It was one of the earliest publicly accessible secondary structure prediction servers, building on earlier methods like PHD and PSIPRED. The server was designed to be simple: users submit a sequence, and within minutes receive a prediction. The initial version used a jury of neural networks trained on a non-redundant set of protein structures. In 2008, Jpred 3 was released, incorporating a more sophisticated algorithm that used a two-stage neural network and improved alignment generation. The latest major version, Jpred 4, launched in 2015, introduced a faster pipeline and updated underlying databases. The server has been continuously maintained, with periodic updates to reflect new protein structure data and algorithmic improvements. ## Methodology Jpred's prediction pipeline begins with a query sequence. It first runs PSI-BLAST to search against a sequence database (e.g., UniProt) to find homologous sequences, which are then aligned to create a multiple sequence alignment. This alignment is converted into a position-specific scoring matrix (PSSM) that captures evolutionary conservation patterns. The PSSM is fed into a neural network, specifically a recurrent neural network with long short-term memory (LSTM) units in Jpred 4, which processes the sequence context. The network outputs probabilities for each residue being in one of three secondary structure states: H (helix), E (strand), or C (coil). A separate network predicts solvent accessibility (buried or exposed). The final output includes a confidence score per residue, typically ranging from 0 to 9. Training data comes from the Protein Data Bank (PDB), filtered to remove redundancy. The neural network is trained using backpropagation and a cross-entropy loss function. The accuracy of Jpred 4 is reported to be around 82% for three-state prediction (Q3), which is competitive with other modern methods. ## Usage and Applications Jpred is used by researchers to annotate protein sequences without known structures, aiding in functional inference and experimental design. It is often a first step in structural bioinformatics pipelines, such as homology modeling and fold recognition. The server also provides a batch mode for analyzing multiple sequences, and an API for programmatic access. Beyond secondary structure, Jpred output can be used to guide site-directed mutagenesis experiments, identify functional residues, and assist in protein engineering. The server has been cited in thousands of scientific publications, making it a standard tool in the field. ## Availability and Impact Jpred is freely available to academic users at the University of Dundee's website. It runs on a cluster of Linux servers, ensuring quick turnaround times. The server has been continuously funded by research grants, including from the Biotechnology and Biological Sciences Research Council (BBSRC). Jpred's impact lies in its accessibility and reliability. It has been a benchmark for secondary structure prediction, and its outputs are often compared with other tools like PSIPRED and JPred's successor, JNet. The server's design has influenced later web-based bioinformatics tools, emphasizing ease of use and clear output visualization. ## Related Tools and Future Directions Jpred is part of a broader ecosystem of protein structure prediction tools, including [artificial-intelligence](https://www.wikiprompt.org/wiki/artificial-intelligence)-based methods like AlphaFold, which predict full 3D structures. While Jpred focuses on secondary structure, its integration of [machine-learning](https://www.wikiprompt.org/wiki/machine-learning) techniques has evolved over time. Future updates may incorporate [deep-learning](https://www.wikiprompt.org/wiki/deep-learning) architectures, such as [transformer](https://www.wikiprompt.org/wiki/transformer) models, to further improve accuracy. The server remains actively maintained, with the Barton Group continuing to develop JNet, the underlying neural network. As of the current date, Jpred 4 is the standard version, and no major new release has been announced, but ongoing improvements to databases and algorithms are expected. --- Source: https://www.wikiprompt.org/wiki/jpred License: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/) Last updated: 2026-09-14T06:31:54.174988+00:00