RaptorX is a web-based server for protein structure modeling that provides secondary structure prediction, template-based tertiary structure modeling, alignment quality assessment, and probabilistic alignment sampling. It distinguishes itself by its ability to generate high-quality structural models for proteins with only remote templates. RaptorX uses a novel nonlinear scoring function and a probabilistic-consistency algorithm to improve alignment quality and model accuracy. It processes sequences of 200 amino acids in ~35 minutes and has processed ~6,000 sequences from ~1,600 users since its release in August 2011. RaptorX is particularly effective in cases where no close homologs are available, as it uses a profile-entropy scoring method and conditional random fields (CRFs) to integrate biological signals into a nonlinear threading score function. It also implements a multiple-template threading (MTT) procedure that allows the use of multiple templates to model a single target sequence, improving alignment accuracy by exploiting inter-template similarity. RaptorX was ranked second in the CASP9 competition, outperforming other servers in generating the best alignments for the 50 most difficult targets. It also provides domain parsing, disorder prediction, and secondary structure prediction, along with a conditional neural field (CNF)-based protocol for determining secondary structure distribution. RaptorX can be used to generate custom pairwise alignments and multiple-template alignments, and it provides a detailed overview of modeling accuracy. RaptorX is not biased toward any specific protein type but has limitations due to insufficient coverage of sequence and structure databases. It can generate reliable models for sequences with low similarity templates in the PDB. The server is compatible with three popular web browsers and requires protein sequences in FASTA format. Users can submit jobs for alignment or structure prediction, and the server provides detailed results, including predicted secondary structure, disorder prediction, domain parsing, and functional annotation. RaptorX is supported by various grants and is used for a wide range of applications, including binding site and epitope prediction, molecular docking, and macromolecular docking. It can also improve multiple-sequence alignments by incorporating structure information from templates. The server is designed to be efficient and user-friendly, with detailed instructions for job submission, monitoring, and result viewing. The time required for processing varies depending on the job type and sequence length, with an average of 30–35 minutes for small protein sequences.RaptorX is a web-based server for protein structure modeling that provides secondary structure prediction, template-based tertiary structure modeling, alignment quality assessment, and probabilistic alignment sampling. It distinguishes itself by its ability to generate high-quality structural models for proteins with only remote templates. RaptorX uses a novel nonlinear scoring function and a probabilistic-consistency algorithm to improve alignment quality and model accuracy. It processes sequences of 200 amino acids in ~35 minutes and has processed ~6,000 sequences from ~1,600 users since its release in August 2011. RaptorX is particularly effective in cases where no close homologs are available, as it uses a profile-entropy scoring method and conditional random fields (CRFs) to integrate biological signals into a nonlinear threading score function. It also implements a multiple-template threading (MTT) procedure that allows the use of multiple templates to model a single target sequence, improving alignment accuracy by exploiting inter-template similarity. RaptorX was ranked second in the CASP9 competition, outperforming other servers in generating the best alignments for the 50 most difficult targets. It also provides domain parsing, disorder prediction, and secondary structure prediction, along with a conditional neural field (CNF)-based protocol for determining secondary structure distribution. RaptorX can be used to generate custom pairwise alignments and multiple-template alignments, and it provides a detailed overview of modeling accuracy. RaptorX is not biased toward any specific protein type but has limitations due to insufficient coverage of sequence and structure databases. It can generate reliable models for sequences with low similarity templates in the PDB. The server is compatible with three popular web browsers and requires protein sequences in FASTA format. Users can submit jobs for alignment or structure prediction, and the server provides detailed results, including predicted secondary structure, disorder prediction, domain parsing, and functional annotation. RaptorX is supported by various grants and is used for a wide range of applications, including binding site and epitope prediction, molecular docking, and macromolecular docking. It can also improve multiple-sequence alignments by incorporating structure information from templates. The server is designed to be efficient and user-friendly, with detailed instructions for job submission, monitoring, and result viewing. The time required for processing varies depending on the job type and sequence length, with an average of 30–35 minutes for small protein sequences.