4. Protein Search, Alignment, & DB Management Tools

Foldseek: A fast structural-alignment and search tool that encodes tertiary amino-acid interactions as a structural alphabet (3Di), letting structure comparisons run with sequence-alignment-speed algorithms. Foldseek is four to five orders of magnitude faster than Dali, TM-align, and CE at comparable sensitivity, making structure-based search practical against databases of hundreds of millions of predicted structures.
Preprint posted: Feb 09, 2022
Paper published: May 08, 2023
Paper | Preprint | Code (GitHub) | Webserver

PS-GO: A parametric protein search engine that integrates protein structure, sequence information, and computable parameters. PS-GO is inspired by the classic Google’s PageRank algorithm to offer a protein search engine to effectively retrieve, compare, analyse, and interpret the growing body of protein data. PS-GO provides an interface that enables users to give specific parameters to search dbs for protein sequences and structures, and optimize/redesign proteins via parametric protein design approach. PS-GO supports range-based parameter search, such as RC.Score, hydrophobicity, instability, size, isoelectric point, and SA. PS-GO also supports natural language queries describing desired protein properties or functions, powered by OpenAI’s GPT-4 (to convert the natural language query into structured parameter conditions). For its full functionality, PS-GO links with PROFASA (Protein Fragment And Structure Analysis), another web-based resource from the same research group.
Paper published: April 8, 2024
Paper | Webserver (PS-GO) | Webserver (PROFASA)

PLMSearch/PLMAlign: A fast protein language model-based method for homologous protein search from sequence input. PLMSearch searches millions of query-target pairs in seconds similar to MMseqs2 but with 3x increase in sensitivity comparable to SOTA structure-based search methods. PLMSearch also captures remote homology pairs similar to structure-based methods. PLMSearch integrates 3 main steps/modules: 1. PfamClan filters for similarity based on Pfam clan domain, 2. SS-predictor predicts structural similarity for all query-target pairs, 3. PLMSearch-PLMAlign provides sequence alignment and scores.
Paper published: March 30, 2024
Paper | Code (GitHub) - PLMSearch | Code (Code Ocean) - PLMSearch | Webserver (PLMSearch) | Code (GitHub) - PLMAlign | Webserver (PLMAlign)

AlphaSync: A computational pipeline and database from St. Jude that updates and synchronizes the AlphaFold protein structure database with UniProt, adding residue-level annotations including buried/core vs. surface classification, backbone dihedrals and secondary-structure categories, inter-residue contacts, solvent-accessible surface area (SASA), and intrinsically disordered region (IDP) prediction.
Paper published: December 2025
Paper | Web Interface

BitterDB: A major update to BitterDB, a database of bitter-tasting molecules and their receptors, from the Niv lab (Hebrew University of Jerusalem). The 2024 update roughly triples its content since 2019: over 2,200 bitter-tasting molecules and close to 1,800 ligand–TAS2R (bitter taste receptor) associations drawn from 253 publications, extended to 66 species and 236 receptors, with 509 new bitter peptides and links to AlphaFold- and cryo-EM-derived receptor structures. The database underlies a family of downstream prediction tools built on its data, including BitterPredict, BitterIntense, BitterMatch, BitterMasS, BitterX, e-Bitter, BitterSweetForest, and VirtualTaste.
Paper published: November 13, 2024
Paper | Database