pith. machine review for the scientific record. sign in

arxiv: 2605.05259 · v1 · submitted 2026-05-06 · 🧬 q-bio.BM · cond-mat.mtrl-sci· cs.AI· q-bio.QM

Recognition: unknown

Enhancing Cryo-EM Density Map Segmentation in Phenix for Improved Atomic Model Building

Chenwei Zhang

Pith reviewed 2026-05-09 16:40 UTC · model grok-4.3

classification 🧬 q-bio.BM cond-mat.mtrl-scics.AIq-bio.QM
keywords PhenixCraftcryo-EMdensity map segmentationAlphaFold integrationatomic model buildingstructural biologyautomated pipelineTM-score
0
0 comments X

The pith

PhenixCraft integrates AlphaFold predictions to improve cryo-EM density map segmentation in Phenix, yielding higher TM-scores and sequence accuracy in atomic models.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces PhenixCraft as an automated pipeline that folds AlphaFold structure predictions into the map-segmentation stage of Phenix. Traditional Phenix segmentation often fails on noisy or artifact-laden cryo-EM maps, leaving gaps that force manual fixes and lower model quality. By supplying predicted atomic coordinates to guide where density belongs to which chain, PhenixCraft produces models with measurably better TM-scores and sequence accuracy. A sympathetic reader would care because cryo-EM is now the dominant route to large protein complexes, and any reliable automation step shortens the time from raw map to usable atomic coordinates.

Core claim

PhenixCraft is a fully automated pipeline that inserts AlphaFold predictions into Phenix’s map-segmentation step to overcome noise and artifacts, thereby producing atomic models with superior TM-scores and sequence accuracy compared with standard Phenix workflows.

What carries the argument

PhenixCraft pipeline, which augments Phenix segmentation by feeding AlphaFold-predicted structures as additional constraints to assign density regions to specific chains or residues.

If this is right

  • Atomic models can be completed with less manual intervention when maps contain moderate noise or missing density.
  • Sequence accuracy improves because AlphaFold guidance helps assign the correct residues to observed density blobs.
  • The pipeline remains fully automated, allowing high-throughput processing of multiple maps without user-guided segmentation.
  • Performance gains are demonstrated on the tested cases through direct comparison of TM-scores and sequence matches.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same segmentation-aid idea could be ported to other modeling packages that currently rely on density-thresholding alone.
  • If the method tolerates moderate AlphaFold errors, it may still work on lower-resolution maps where experimental density is ambiguous.
  • A useful next test would be to quantify how much the gain shrinks when AlphaFold predictions come from more distant homologs rather than close relatives.

Load-bearing premise

AlphaFold predictions are sufficiently accurate that their use in segmentation does not introduce new placement errors or biases into the final atomic model.

What would settle it

A side-by-side test on a benchmark set of cryo-EM maps with deposited ground-truth models, measuring whether PhenixCraft models show statistically higher TM-scores and sequence identity than models built with unmodified Phenix on the same maps.

Figures

Figures reproduced from arXiv: 2605.05259 by Chenwei Zhang.

Figure 1
Figure 1. Figure 1: The architecture of the PhenixCraft pipeline. All maps are shown in rectangular boxes with pale orange view at source ↗
Figure 2
Figure 2. Figure 2: Evaluation of the built models against the reference PDB structure for PhenixCraft and view at source ↗
Figure 3
Figure 3. Figure 3: Constructed models by PhenixCraft and phenix.map_to_model. Chains are colored separately. The TM-scores for each model are listed alongside. a. ERAD-associated E3 ubiquitin-protein ligase HRD1 (EMDB ID: 8637; PDB ID: 5V6P; reported resolution: 4.1 Å). b. Rotavirus VP6 (EMDB ID: 6272; PDB ID: 3J9S; reported resolution: 2.6 Å). c. Beta-galactosidase in complex with a cell-permeant inhibitor (EMDB ID: 2984; P… view at source ↗
read the original abstract

We introduce PhenixCraft, a fully automated pipeline for building atomic models from cryo-EM density maps. By integrating AlphaFold predictions, we enhance the map-segmentation step in Phenix during model building, addressing challenges posed by noise and artifacts that traditionally hinder this step. Our results demonstrate PhenixCraft's superior performance in TM-scores and sequence accuracy, significantly improving upon the limitations and inefficiencies of traditional model building using Phenix.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 0 minor

Summary. The manuscript introduces PhenixCraft, a fully automated pipeline for building atomic models from cryo-EM density maps. It integrates AlphaFold predictions to enhance the map-segmentation step in Phenix, claiming to address noise and artifacts and demonstrating superior TM-scores and sequence accuracy over traditional Phenix model building.

Significance. If the performance improvements hold under rigorous validation, this work has the potential to enhance the efficiency and accuracy of atomic model building in cryo-EM structural biology by combining deep learning-based predictions with established software tools like Phenix.

major comments (3)
  1. Abstract: The abstract asserts superior TM-scores and sequence accuracy but provides no quantitative values, test datasets, error bars, baseline comparisons, or statistical tests, making the central performance claim impossible to evaluate from the given information.
  2. Methods section: The integration of AlphaFold predictions into the Phenix segmentation step is not described in sufficient detail, including the exact mechanism of incorporation and any safeguards against propagating inaccuracies from AlphaFold models (e.g., in low-confidence loop regions or novel folds).
  3. Results section: No per-region breakdown or control experiments are presented to show that AlphaFold integration improves segmentation without introducing new biases, which is critical given that the segmentation step is identified as the bottleneck.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their thoughtful review and constructive comments on our manuscript. We have carefully considered each point and provide detailed responses below, along with plans for revisions to address the concerns raised.

read point-by-point responses

  1. Referee: Abstract: The abstract asserts superior TM-scores and sequence accuracy but provides no quantitative values, test datasets, error bars, baseline comparisons, or statistical tests, making the central performance claim impossible to evaluate from the given information.

    Authors: We agree with this observation. The abstract in the current version is indeed concise and lacks specific quantitative details. In the revised manuscript, we will update the abstract to include key quantitative results, such as the TM-score improvements and sequence accuracy metrics with error bars, the test datasets used, baseline comparisons to standard Phenix, and mention of the statistical tests applied. This will make the performance claims evaluable from the abstract. revision: yes

  2. Referee: Methods section: The integration of AlphaFold predictions into the Phenix segmentation step is not described in sufficient detail, including the exact mechanism of incorporation and any safeguards against propagating inaccuracies from AlphaFold models (e.g., in low-confidence loop regions or novel folds).

    Authors: We will revise the Methods section to provide sufficient detail on the integration of AlphaFold predictions into the Phenix segmentation step. This will include the exact mechanism of incorporation as well as safeguards against propagating inaccuracies from AlphaFold models, particularly in low-confidence loop regions or for novel folds. revision: yes

  3. Referee: Results section: No per-region breakdown or control experiments are presented to show that AlphaFold integration improves segmentation without introducing new biases, which is critical given that the segmentation step is identified as the bottleneck.

    Authors: We agree that additional analyses are needed. In the revised manuscript, we will present per-region breakdowns in the Results section and include control experiments to confirm that the AlphaFold integration improves segmentation without introducing new biases. This is particularly important as the segmentation step is the identified bottleneck. revision: yes

Circularity Check

0 steps flagged

No circularity: empirical method comparison with external inputs

full rationale

The paper introduces PhenixCraft as a pipeline that feeds AlphaFold predictions into Phenix's map-segmentation step and reports higher TM-scores plus sequence accuracy versus baseline Phenix on test maps. No equations, fitted parameters, uniqueness theorems, or derivation chain exist. AlphaFold is treated as an independent black-box input whose outputs are used as additional restraints; performance numbers are measured against external ground-truth models rather than being forced by the method's own definitions or self-citations. This is a standard applied-methods paper whose central claim is falsifiable by independent benchmarks and does not reduce to its inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim depends on the untested premise that AlphaFold outputs are sufficiently reliable to guide Phenix segmentation on noisy maps; no free parameters or invented physical entities are described in the abstract.

axioms (1)
  • domain assumption AlphaFold structure predictions are accurate enough to improve density map segmentation in Phenix
    The method explicitly integrates AlphaFold predictions as the key enhancement step.
invented entities (1)
  • PhenixCraft pipeline no independent evidence
    purpose: Automated atomic model building from cryo-EM maps via enhanced segmentation
    New named software pipeline introduced in the abstract.

pith-pipeline@v0.9.0 · 5367 in / 1359 out tokens · 26902 ms · 2026-05-09T16:40:10.259427+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

26 extracted references · 2 canonical work pages

  1. [1]

    The resolution revolution.Science, 343(6178):1443–1444, 2014

    Werner Kühlbrandt. The resolution revolution.Science, 343(6178):1443–1444, 2014

    2014

  2. [2]

    Overview and future of single particle electron cryomicroscopy.Archives of biochemistry and biophysics, 581:19–24, 2015

    Richard Henderson. Overview and future of single particle electron cryomicroscopy.Archives of biochemistry and biophysics, 581:19–24, 2015

    2015

  3. [3]

    Cryo-em: the resolution revolution and drug discovery.SLAS Discovery: Advancing the Science of Drug Discovery, 26(1):17–31, 2021

    Taiana Maia de Oliveira, Lotte van Beek, Fiona Shilliday, Judit É Debreczeni, and Chris Phillips. Cryo-em: the resolution revolution and drug discovery.SLAS Discovery: Advancing the Science of Drug Discovery, 26(1):17–31, 2021

    2021

  4. [4]

    Chenwei Zhang, Anne Condon, and Khanh Dao Duc. A comprehensive survey and benchmark of deep learning- based methods for atomic model building from cryo-electron microscopy density maps.Briefings in Bioinformatics, 26(4):bbaf322, 2025

    2025

  5. [5]

    Novel artificial intelligence-based approaches for ab initio structure determination and atomic model building for cryo-electron microscopy.Micromachines, 14(9):1674, 2023

    Megan C DiIorio and Arkadiusz W Kulczyk. Novel artificial intelligence-based approaches for ab initio structure determination and atomic model building for cryo-electron microscopy.Micromachines, 14(9):1674, 2023

    2023

  6. [6]

    Alignment of partially overlapping cryo-em maps using unbalanced gromov-wasserstein divergence.PRX Life, 3(2):023003, 2025

    Aryan Tajmir Riahi, Chenwei Zhang, Anne Condon, James Chen, and Khanh Dao Duc. Alignment of partially overlapping cryo-em maps using unbalanced gromov-wasserstein divergence.PRX Life, 3(2):023003, 2025

    2025

  7. [7]

    A fully automatic method yielding initial models from high-resolution cryo-electron microscopy maps.Nature methods, 15(11):905–908, 2018

    Thomas C Terwilliger, Paul D Adams, Pavel V Afonine, and Oleg V Sobolev. A fully automatic method yielding initial models from high-resolution cryo-electron microscopy maps.Nature methods, 15(11):905–908, 2018

    2018

  8. [8]

    A graph neural network approach to automated model building in cryo-em maps

    Kiarash Jamali, Dari Kimanius, and Sjors HW Scheres. A graph neural network approach to automated model building in cryo-em maps. InThe Eleventh International Conference on Learning Representations, 2022

    2022

  9. [9]

    Jonas Pfab, Nhut Minh Phan, and Dong Si. Deeptracer for fast de novo cryo-em protein structure modeling and special studies on cov-related complexes.Proceedings of the National Academy of Sciences, 118(2):e2017525118, 2021

    2021

  10. [10]

    Modeling symmetric macromolecular structures in rosetta3.PloS one, 6(6):e20450, 2011

    Frank DiMaio, Andrew Leaver-Fay, Phil Bradley, David Baker, and Ingemar André. Modeling symmetric macromolecular structures in rosetta3.PloS one, 6(6):e20450, 2011

    2011

  11. [11]

    Highly accurate protein structure prediction with alphafold.Nature, 596(7873):583–589, 2021

    John Jumper, Richard Evans, Alexander Pritzel, Tim Green, Michael Figurnov, Olaf Ronneberger, Kathryn Tunyasuvunakool, Russ Bates, Augustin Žídek, Anna Potapenko, et al. Highly accurate protein structure prediction with alphafold.Nature, 596(7873):583–589, 2021

    2021

  12. [12]

    Auto-sharpening cryo-em or crystallographic maps with auto_sharpen

    Tom Terwilliger et al. Auto-sharpening cryo-em or crystallographic maps with auto_sharpen. https: //phenix-online.org/documentation/reference/auto_sharpen.html. Accessed: 2024

    2024

  13. [13]

    Automated map sharpening by maximization of detail and connectivity.Acta Crystallographica Section D: Structural Biology, 74(6):545–559, 2018

    Thomas C Terwilliger, Oleg V Sobolev, Pavel V Afonine, and Paul D Adams. Automated map sharpening by maximization of detail and connectivity.Acta Crystallographica Section D: Structural Biology, 74(6):545–559, 2018

    2018

  14. [14]

    Improvement of cryo-em maps by simultaneous local and non-local deep learning.Nature communications, 14(1):3217, 2023

    Jiahua He, Tao Li, and Sheng-You Huang. Improvement of cryo-em maps by simultaneous local and non-local deep learning.Nature communications, 14(1):3217, 2023

    2023

  15. [15]

    Cryosamu: Enhancing 3d cryo-em density maps of protein structures at intermediate resolution with structure-aware multimodal u-nets.arXiv preprint arXiv:2503.20291, 2025

    Chenwei Zhang and Khanh Dao Duc. Cryosamu: Enhancing 3d cryo-em density maps of protein structures at intermediate resolution with structure-aware multimodal u-nets.arXiv preprint arXiv:2503.20291, 2025

    work page arXiv 2025

  16. [16]

    Deepemhancer: a deep learning solution for cryo-em volume post-processing.Communications biology, 4(1):874, 2021

    Ruben Sanchez-Garcia, Josue Gomez-Blanco, Ana Cuervo, Jose Maria Carazo, Carlos Oscar S Sorzano, and Javier Vargas. Deepemhancer: a deep learning solution for cryo-em volume post-processing.Communications biology, 4(1):874, 2021

    2021

  17. [17]

    Struc2mapgan: improving synthetic cryogenic electron microscopy density maps with generative adversarial networks.Bioinformatics Advances, 5(1):vbaf179, 2025

    Chenwei Zhang, Anne Condon, and Khanh Dao Duc. Struc2mapgan: improving synthetic cryogenic electron microscopy density maps with generative adversarial networks.Bioinformatics Advances, 5(1):vbaf179, 2025

    2025

  18. [18]

    T. C. Terwilliger. Computational crystallography newsletter.Comput. Crystallogr. Newsl., 9:51–57, 2018

    2018

  19. [19]

    phenix.map_box: extract box with model and map around selected atoms

    Tom Terwilliger et al. phenix.map_box: extract box with model and map around selected atoms. https: //phenix-online.org/documentation/reference/map_box.html. Accessed: 2024

    2024

  20. [20]

    Emdatabank unified data resource for 3dem

    Catherine L Lawson, Ardan Patwardhan, Matthew L Baker, Corey Hryc, Eduardo Sanz Garcia, Brian P Hudson, Ingvar Lagerstedt, Steven J Ludtke, Grigore Pintilie, Raul Sala, et al. Emdatabank unified data resource for 3dem. Nucleic acids research, 44(D1):D396–D403, 2016

    2016

  21. [21]

    The protein data bank.Acta Crystallographica Section D: Biological Crystallography, 58(6):899–907, 2002

    Helen M Berman, Tammy Battistuz, Talapady N Bhat, Wolfgang F Bluhm, Philip E Bourne, Kyle Burkhardt, Zukang Feng, Gary L Gilliland, Lisa Iype, Shri Jain, et al. The protein data bank.Acta Crystallographica Section D: Biological Crystallography, 58(6):899–907, 2002. 7 PhenixCraft

    2002

  22. [22]

    Terwilliger

    Thomas C. Terwilliger. Phenix.map to model: A fully automatic method yielding initial models from high- resolution electron cryo-microscopy maps. https://phenix-online.org/phenix_data/terwilliger/ map_to_model_2018/, 2018

    2018

  23. [23]

    Colabfold: making protein folding accessible to all.Nature methods, 19(6):679–682, 2022

    Milot Mirdita, Konstantin Schütze, Yoshitaka Moriwaki, Lim Heo, Sergey Ovchinnikov, and Martin Steinegger. Colabfold: making protein folding accessible to all.Nature methods, 19(6):679–682, 2022

    2022

  24. [24]

    Ucsf chimerax: Structure visualization for researchers, educators, and developers

    Eric F Pettersen, Thomas D Goddard, Conrad C Huang, Elaine C Meng, Gregory S Couch, Tristan I Croll, John H Morris, and Thomas E Ferrin. Ucsf chimerax: Structure visualization for researchers, educators, and developers. Protein Science, 30(1):70–82, 2021

    2021

  25. [25]

    Scoring function for automated assessment of protein structure template quality

    Yang Zhang and Jeffrey Skolnick. Scoring function for automated assessment of protein structure template quality. Proteins: Structure, Function, and Bioinformatics, 57(4):702–710, 2004

    2004

  26. [26]

    Mm-align: a quick algorithm for aligning multiple-chain protein complex structures using iterative dynamic programming.Nucleic acids research, 37(11):e83–e83, 2009

    Srayanta Mukherjee and Yang Zhang. Mm-align: a quick algorithm for aligning multiple-chain protein complex structures using iterative dynamic programming.Nucleic acids research, 37(11):e83–e83, 2009. 8 PhenixCraft Supplementary Material EMDB: 8782 PDB: 5W81TM-score: 0.30 a b c TM-score: 0.18 Figure S1: In (a), the experimental density map is colored blue....

    work page arXiv 2009