LOMETS (Local Meta-Threading Server, version 3) is a new generation of meta-server approach
to template-based protein structure prediction and structure-based function annotation, which integrates multiple
deep learning-based threading methods (CEthreader, DisCovER, EigenThreader, Hybrid-CEthreader, MapAlign)
and state-of-the-art profile-based programs (FFAS3D, HHpred, HHsearch, MRFsearch, MUSTER, SparksX).
To model sequences without homologous templates, an L-BFGS folding system is introduced to
construct full-length models from deep-learning contact/distance restraints by DeepPotential and LOMETS top templates.
Large-scaled benchmark tests showed that the overall template-recognition
accuracy is significantly beyond its predecessors (LOMETS and LOMETS2), due to the integration
of deep-learning techniques.
LOMETS3 participated in
as 'Zhang-TBM' and was ranked as one of the top methods for automatic protein structure prediction.
A detailed description of the LOMETS3 server can be seen on
the About LOMETS page.
Please post your questions and comments about LOMETS at the
Service System Discussion Board.
LOMETS has been updated to LOMETS3 with major updates, including:
Template library: While template libraries in former LOMETS are generated separately for different
threading programs, which can result in inconsistent update and completeness of template
structures, a unified and comprehensive template library is now created and weekly
updated for all threading programs.
MSA profile: A deep multiple sequence alignment (DeepMSA)
approach is developed to create deep sequence profiles from metagenome sequence databases for
all template proteins, which significantly improves the accuracy of
almost all the profile- and deep learning-based threading alignments.
Threading programs: More than half of the old threading programs were renewed and/or
replaced by the state-of-the-art methods, including those combining the cutting-edge deep-learning techniques.
Re-ranking method: Residue-Residue distances, contacts, and hydrogen bond geometries are predicted from DeepPotential. A new scoring function, which combines residue distances, contacts, hydrogen bonds, and a profile score, is used to re-rank the templates for profile-based threadings.
Ab initio structure modeling: An L-BFGS system is introduced to construct full-length structure models
for non-homologous target sequences based on spatial restraints predicted by DeepPotential and those deduced
from top threading templates.
Atomic model refinement: New refinement pipeline based on FG-MD and
FASPR is used to refine and re-pack the side-chain conformation of the final models.
Structural analogs: TM-align is used to search the first LOMETS3 model through all structures
in the PDB library, where the top 10 protein structures with the closest structural similarity,
i.e., the highest TM-score, to the target are reported.
Functional annotations: Completely redesigned output page, which now contains
structure-based function annotations (including Gene Ontology term, Enzyme Commission number, and Ligand Binding residues) derived
from threading templates.
of the updated LOMETS server includes
- Secondary structure prediction
- Solvent accessibility prediction
- Contact-map and distance-map prediction by DeepPotential
- The best ten threading templates selected from 110 (=11x10) templates
- Full-length models built by MODELLER based on the top-five selected templates for homologous targets,
or built by an L-BFGS system using distance restraints from DeepPotential and LOMETS templates for non-homologous targets.
- The best ten similar structure identified by TM-align using the first LOMETS model, and the associated functional annotations.
- Functional annotations (Gene Ontology term, Enzyme Commission number, and Ligand Binding residues) derived from top-ranking threading templates
Tips for modeling multi-domain proteins which are usually have >500 AA:
- Use FUpred, ThreaDom or ThreaDomEx servers to predict the domain boundary for your sequence, and
then split the full-length sequence into domain-level sequences by FUpred, ThreaDom or ThreaDomEx domain partition information.
- Submit the domain-level sequences to LOMETS server to get the domain-level models, respectively.
- Submit the full-length sequence and domain-level models to DEMO server, which will automatically assemble the full-length model.
(You can also directly submit the full-length sequence to LOMETS server to get the model. Next, manually determine domain
partitions based on the structure model of LOMETS as structural models generally show clear domain boundaries.
Finally, repeat Step 2 and 3.)
[Check Previous Jobs]
Wei Zheng, Azam Hussain, Qiqige Wuyun, Robin Pearce, Yang Li, Yang Zhang.
LOMETS3: Integrating deep-learning and profile-alignment for advanced protein template recognition and function annotation,
in preparation, 2020.
Wei Zheng, Chengxin Zhang, Qiqige Wuyun, Robin Pearce, Yang Li, Yang Zhang.
LOMETS2: improved meta-threading server for fold-recognition and structure-based
function annotation for distant-homology proteins.
Nucleic Acids Research, 47: W429-W436 (2019)
[PDF of manuscript]
[PDF of Support Information].
- Sitao Wu, Yang Zhang.
LOMETS: A local meta-threading-server for protein structure prediction.
Nucleic Acids Research, 35: 3375-3382 (2007)
[PDF of manuscript].