Research62

TTPD: Invisible String

We talk about ANOTHER big travel month, federal fights over science funding, preprints, and the intersection of AI and scientific discoveries.

How Silent Code Updates Can Alter Protein Predictions

Software updates can alter protein structure predictions even when model weights stay fixed. A Protenix case study shows why reproducibility requires tracking the entire prediction pipeline.

Don't just do something, sit there: Mindful of the MD Model

Proteins can "shift in their seats" depending on where they sit in the lattice.

Varying crystallization conditions to study lattice effects on diffuse scattering

My summer research project in the Ando Lab

The Tortured Proteins Department: Introducing Prism

Podcast introducing Prism

The Vision for Prism

Charting the course for dynamic structural biology.

qFitLatent: Learning the Motions of Multiconformers

Learning a latent representation of qFit multiconformers to capture the local dynamics that crystallography usually discards.

WaterFlow: Predicting Ordered Water Molecule Positions on Protein Structures

A flow-matching model for predicting ordered water positions on protein structures, and what it reveals about the limits of the data we train it on.

A Year (plus) of DiffUSE

Science progress over the first year-ish of the DiffUSE Project

Coffee is for Closers: making Mac1 diffuse scattering reproducible

diffuse scattering reproducible, yay!

The Tortured Proteins Department, Episode 17

The Tortured Proteins Department Podcast, Episode 17

WaterFlow: Prediction of Ordered Water Molecule Positions on Protein Structures

Opens on biorxiv.org/content/10.64898/2026.08.26.747373v1

A flow-matching generator and confidence model for predicting the positions of ordered water molecules in protein structures. WaterFlow outperforms the existing state of the art at every precision level, and its novel predictions often fall on positive difference density, placing waters at sites the original depositions omitted. Preprint, bioRxiv. DOI: 10.64898/2026.08.26.747373

Revealing hidden signals in raw X-ray crystallography data with mdx2

Mdx2 is a powerful and open-source toolkit for diffuse data processing.

Modeling the contribution of harmonic lattice vibrations to the diffuse signal with GOODVIBES

GOODVIBES allows users to model the contribution of harmonic lattice vibrations to diffuse scattering.

Recovering Conformational Heterogeneity from the Protein Data Bank at Scale

Opens on thestacks.org/publications/qfit-at-scale

An ensemble dataset of over 60,000 structures. Applying qFit to re-refined PDB entries better than 2 Å to recover latent conformational heterogeneity at scale — the largest dataset of experimentally derived ensemble models. DOI: 10.82153/pff0-ck46

Displaying Conformational Heterogeneity in BANDICOOT

In preparation for hierarchical hetereogeneity encoding, BANDICOOT has a new, more intuitive display mode for alternate conformations.

mmCIF Explorer and the Heterogeneity Proposal

A playground for the PDBx/mmCIF file exploration, a look at the current proposal for encoding correlated heterogeneity

Is it all just the lattice?

Taylor's update on his work in the diffUSE project and the roadmap to a complete forward model of diffuse scattering.

Cruel Kinase Summer, Episode 16

The Tortured Proteins Department Podcast, Cruel Kinase Summer

BANDICOOT: Coot 0.9 That Runs on MacOS Tahoe

Reviving Coot 0.9 for MacOS Tahoe as Bandicoot, a bespoke and integrated molecular modeling environment.

Introducing PLUG: A Compiler for Leakage-free Protein Function Benchmarking

Leakage-free Training to Predict Protein Function

macromolecular-observables

Opens on github.com/prism-science/macromolecular-observables

A package for conveying and transforming information about proteins — ensembles, structures, and measurements. A joint project of Radial Science and Reciprocal Spaceship.

Supplying a user-defined bulk solvent map for refinement

Opens on thestacks.org/publications/method-bulksolvent

A patch to Phenix refinement that lets users supply their own bulk solvent map in place of the default flat binary mask, opening refinement to MD, 3D-RISM and neural-network solvent models. DOI: 10.82153/2me0-hd96

The Tortured Proteins Department, Episode 15

The Tortured Proteins Department Podcast, Episode 15

DiffUSE Open Science Policies and Guidelines

Astera’s open science policy is that all work products created with Astera dollars must be made available under CC-BY (attribution required, recommended) or CC 0 (completely free use). Even if it’s…

Open Development for Biomolecular Ensemble Generation

The diffUSE Project releases *sampleworks*, moving beyond open science to science in the open

sampleworks: A Modular Platform for Experimentally Guided Biomolecular Ensemble Generation

Opens on thestacks.org/publications/sampleworks-release

A modular framework for generating and evaluating biomolecular conformational ensembles from structure predictors guided by experimental data, benchmarked across Boltz-2, Protenix and RosettaFold3. DOI: 10.82153/jkxj-tw08

consensus_water_analysis

Opens on github.com/prism-science/consensus_water_analysis

Analysis of consensus ordered water across Protein Data Bank depositions.

qFitLatent

Opens on github.com/prism-science/qFitLatent

A model to learn qFit multiconformer dynamics with an invariant backbone.

The Tortured Proteins Department, Episode 14

The Tortured Proteins Department Podcast, Episode 14

pdbx_hierarchy

Opens on github.com/prism-science/pdbx_hierarchy

Tools for working with hierarchical heterogeneity mmCIF models.

Encoding Heterogeneity in mmCIF Files

Explicitly encoding hierarchical heterogeneity in mmCIF files

mmcif_encoding

Opens on github.com/prism-science/mmcif_encoding

Examples and descriptions related to mmCIF encoding of hierarchical heterogeneity.

Water analysis

Case analysis of consensus water in PDB

md-workflows

Opens on github.com/prism-science/md-workflows

Molecular dynamics workflows for simulating diffuse scattering.

goodvibes

Opens on github.com/prism-science/goodvibes

Lattice vibration model for diffuse scattering

The Tortured Proteins Department, Episode 13

The Tortured Proteins Department Podcast, Episode 13

mdx2-workflows

Opens on github.com/prism-science/mdx2-workflows

Workflows for diffuse scattering data reduction with mdx2.

Expanding The DiffUSE Project

Expanding The DiffUSE Project

To ensembles and beyond!

Sampleworks is a framework for integrating structural biology observables, structure predictors, and guidance to improve the modeling of conformational ensembles.

Shake It Up!

MD simulations of changes in diffuse scattering depending on ligand binding

DiffUSE January 2026 Retreat: From Coast to Coast, Diffuse Scattering Reproduces

A report on our January 2026 all-hands meeting

WaterFlow

Opens on github.com/prism-science/WaterFlow

Predicting water molecule placements on protein surfaces using flow matching conditioned on learned protein structure embeddings.

In the Cloud

Next steps in diffUSE MD simulations

Quarterly All-Hands Meeting Summary

A report on our October 2025 all-hands meeting

3-2-1 Contact: Comparing MD simulations with diffuse data

Insights from the first comparisons of MD simulations to diffuse data

Ensemble <-> Multiconformer Model Conversion

Converting between ensemble and multiconformer models

Optimizing Molecular Dynamics Weights with Machine Learning Tools

Using gradient-based optimization to identify the most physically relevant portions of MD trajectories by maximizing agreement with diffuse scattering data.

sampleworks

Opens on github.com/prism-science/sampleworks

Framework for modified sampling from biomolecular generative models

Taylor's Diffuse Rotation

What Taylor got done during his rotation.

Sample Shipping

Sample preparation for future beamtimes

The Barrier of Barrier Materials

It is hard to have happy crystals and low background at the same time

Next Steps with the Macrodomain Data

What to do next about improving macrodomain data quality and processing

From systems operators to systems architects

Opens on seemay.substack.com/p/from-systems-operators-to-systems

Going up a level from data generation to think about the data systems we design and embed

Encoding data for a protein dynamics centered AlphaFold

Re-engineer the encoding and infrastructure of structural biology to embrace dynamics

Modeling Protein Dynamics

Modeling by learning directly from experimental data

Join diffUSE!

Open Jobs at diffUSE

Let's dance!

Molecular-dynamics simulations of diffuse scattering

Getting our feet wet

Data collection at ALS beamline 8.3.1 on June 24, 2025

The DaVinci Dude Dance

Calculated diffraction patterns from different models

Embrace the Mess

A trio of ligand-protein structure papers

DiffUSE Logbook

Opens on diffuse.science/logbook/

An experiment in open science to catalog X-ray datasets, collect notes from synchrotron data collection, share reports from data processing, and post preliminary analysis without filters.