Automated Model Building with ModelCraft

Paul Bond; Jordan Dialpuri; Kathryn Douglas Cowtan

Automated Model Building with ModelCraft

Paul Bond^*, Jordan Dialpuri, Kathryn Douglas Cowtan

^*Corresponding author for this work

Chemistry

Research output: Contribution to conference › Poster

Abstract

ModelCraft is an automated model-building pipeline for X-ray crystallography and cryo-EM that combines Buccaneer for building protein, Nautilus for building DNA/RNA and Refmac/Servalcat for refinement. The X-ray pipeline also includes shift-field refinement, machine-learned pruning of incorrect protein residues, classical density modification, dummy-atom refinement, addition of water, and final rebuilding of side chains. These extra steps make the pipeline more likely to build a complete model when starting from a poor molecular-replacement solution than the previous Buccaneer or Nautilus pipelines. New developments to ModelCraft make it better at building protein/nucleotide complexes. A new program, NucleoFind, uses a deep neural network to predict locations of nucleotide phosphates, sugars and bases. Additionally, deep neural networks have been trained to predict next and previous residue positions and to predict high resolution calculated backbone density from low-resolution maps.

Original language	English
Publication status	Published - 21 Jul 2024
Event	Diffraction Methods in Structural Biology 2024: The Measurement versus the Experiment - Hernack Haus, Berlin, Germany Duration: 21 Jul 2024 → 27 Jul 2024 https://events.gwdg.de/event/650/

Conference

Conference	Diffraction Methods in Structural Biology 2024
Country/Territory	Germany
City	Berlin
Period	21/07/24 → 27/07/24
Internet address	https://events.gwdg.de/event/650/

Cite this

@conference{4929e2a183c141e69a5e7efe901efa9e,

title = "Automated Model Building with ModelCraft",

abstract = "ModelCraft is an automated model-building pipeline for X-ray crystallography and cryo-EM that combines Buccaneer for building protein, Nautilus for building DNA/RNA and Refmac/Servalcat for refinement. The X-ray pipeline also includes shift-field refinement, machine-learned pruning of incorrect protein residues, classical density modification, dummy-atom refinement, addition of water, and final rebuilding of side chains. These extra steps make the pipeline more likely to build a complete model when starting from a poor molecular-replacement solution than the previous Buccaneer or Nautilus pipelines. New developments to ModelCraft make it better at building protein/nucleotide complexes. A new program, NucleoFind, uses a deep neural network to predict locations of nucleotide phosphates, sugars and bases. Additionally, deep neural networks have been trained to predict next and previous residue positions and to predict high resolution calculated backbone density from low-resolution maps.",

author = "Paul Bond and Jordan Dialpuri and Cowtan, {Kathryn Douglas}",

year = "2024",

month = jul,

day = "21",

language = "English",

note = "Diffraction Methods in Structural Biology 2024 : The Measurement versus the Experiment ; Conference date: 21-07-2024 Through 27-07-2024",

url = "https://events.gwdg.de/event/650/",

}

TY - CONF

T1 - Automated Model Building with ModelCraft

AU - Bond, Paul

AU - Dialpuri, Jordan

AU - Cowtan, Kathryn Douglas

PY - 2024/7/21

Y1 - 2024/7/21

N2 - ModelCraft is an automated model-building pipeline for X-ray crystallography and cryo-EM that combines Buccaneer for building protein, Nautilus for building DNA/RNA and Refmac/Servalcat for refinement. The X-ray pipeline also includes shift-field refinement, machine-learned pruning of incorrect protein residues, classical density modification, dummy-atom refinement, addition of water, and final rebuilding of side chains. These extra steps make the pipeline more likely to build a complete model when starting from a poor molecular-replacement solution than the previous Buccaneer or Nautilus pipelines. New developments to ModelCraft make it better at building protein/nucleotide complexes. A new program, NucleoFind, uses a deep neural network to predict locations of nucleotide phosphates, sugars and bases. Additionally, deep neural networks have been trained to predict next and previous residue positions and to predict high resolution calculated backbone density from low-resolution maps.

AB - ModelCraft is an automated model-building pipeline for X-ray crystallography and cryo-EM that combines Buccaneer for building protein, Nautilus for building DNA/RNA and Refmac/Servalcat for refinement. The X-ray pipeline also includes shift-field refinement, machine-learned pruning of incorrect protein residues, classical density modification, dummy-atom refinement, addition of water, and final rebuilding of side chains. These extra steps make the pipeline more likely to build a complete model when starting from a poor molecular-replacement solution than the previous Buccaneer or Nautilus pipelines. New developments to ModelCraft make it better at building protein/nucleotide complexes. A new program, NucleoFind, uses a deep neural network to predict locations of nucleotide phosphates, sugars and bases. Additionally, deep neural networks have been trained to predict next and previous residue positions and to predict high resolution calculated backbone density from low-resolution maps.

M3 - Poster

T2 - Diffraction Methods in Structural Biology 2024

Y2 - 21 July 2024 through 27 July 2024

ER -