Softwares

AlphaFold 3

AlphaFold 3 is the latest version of DeepMind’s AI model for predicting biomolecular structures. Unlike AlphaFold 2, it can model not just proteins, but also complexes with DNA, RNA, and small molecules. It represents a major leap forward in structural biology and drug discovery.

View on Github

AlphaPullDown

AlphaPulldown is a customized implementation of AlphaFold-Multimer designed for customizable high-throughput screening of protein-protein interactions. It extends AlphaFold’s capabilities by incorporating additional run options, such as customizable multimeric structural templates (TrueMultimer), MMseqs2 multiple sequence alignment (MSA) via ColabFold databases, protein fragment predictions, and the ability to incorporate mass spec data as an input using AlphaLink2.

View on Github

BindCraft

BindCraft is a computational tool designed to predict and model protein–ligand interactions. It integrates structural data and AI-driven methods to identify binding sites and estimate binding affinities. BindCraft is useful for drug discovery, molecular docking, and interaction analysis.

View on Github

Boltz-2

Boltz-2 is a next-generation deep learning model for predicting biomolecular structures and binding affinities. It surpasses AlphaFold3-level accuracy by jointly modeling complex interactions, including docking and affinity estimation. As the first model to rival physics-based FEP methods while being 1000x faster, Boltz-2 enables practical and accurate in silico drug screening.

View on Github

Chai-1

Chai-1 is a deep learning model for predicting RNA interaction sites at high resolution. It supports a wide range of partners, including proteins, DNA, and small molecules. Chai-1 enables detailed mapping of biomolecular interfaces for functional analysis and design.

View on Github

ColabFold

ColabFold is an easy-to-use platform for predicting protein structures and complexes using AlphaFold2 and AlphaFold2-Multimer. It streamlines the workflow by automatically generating sequence alignments and templates with MMseqs2 and HHsearch.

View on Github

ProteinMPNN

ProteinMPNN is a deep learning model for protein sequence design based on fixed backbone structures. It generates amino acid sequences that are likely to fold into a given 3D structure. ProteinMPNN is widely used for protein engineering and de novo protein design.

View on Github

RFdiffusion

RFdiffusion is a deep generative model for protein design based on diffusion probabilistic modeling. It builds 3D protein structures from scratch or around functional motifs with high flexibility and control. RFdiffusion is a powerful tool for creating novel proteins with desired shapes and functions.

View on Github

HHalign-Kbest

HHalign-Kbest is useful to automatically obtain optimized alignments and models in case of low sequence identity (<35%) between a query and a template protein. It can generate k suboptimal (e.g. top-k scoring) alignments rather than only the optimal one which may contain small to large errors.

Yu J, Picord G, Tufféry P, Guerois R.
HHalign.KBest: exploring sub-optimal alignments for remote homology comparative modeling
submitted.
Download

iSuperpose

iSuperpose performs the 3D superposition of protein structures by best superimposing the alpha-carbons (or the backbone) of the proteins given a alignment specifying the correspondence between the structures. If no alignment is provided, a structural alignment will be calculated using TMalign. One the alignement is identified, the superposition is achieved using a quaternion based procedure using a specific eigen value calculation implementation. See QBestFit.

Maupetit J, Tufféry P.
Download

XmMol

XmMol is a desktop macromolecular visualization and modeling tool designed to be easy to use, configure and enhance. Its graphics are based on X11, and part of its user interface is based on Motif. Thus it provides a way of displaying structures on any X11 server. Its main features are:

  • interactive graphics of macromolecules on any X11 display. Drawings are performed as wireframes to preserve interactivity. Space filling static images can be obtained by using an interface to external rendering programs such as MolScript and Raster3D.
  • strong ability to be interfaced with external programs. A communication protocol allows XmMol to fork external programs called "delegates" and exchange information. This feature allows the easy implementation of new features of XmMol. This offers possibilities for automatic script execution, new file format I/O implementation, file coordinate modification, implementing external renderers, .... Thus, XmMol can also be used as a graphic debugger for numerical methods applied to molecules (minimizers, ...). Examples of how to implement molecular superimposition, dynamic trajectory animation, as well as calls to external standard programs such as babel or hbplus are provided.
  • Some modelling tools are supported, such as docking facilities, interactive backbone deformation (part of the Forme package). However, the aim of XmMol is mostly to give each user the opportunity to interface its own methods.
Tufféry P.
XmMol: an X11 and motif program for macromolecular visualization and modeling.
J Mol Graph. 1995 Feb;13(1):67-72, 62.
Download

Python libraries

af_analysis Python package for the analysis of AlphaFold protein structure predictions. View on Github
docking_py Python library allowing a simplified use of the Smina, vina, qvina2 and qvinaw docking software. View on Github
Fasta Python library to manage single and multi fasta sequences (Python 2 version). View on Gitlab
Fasta3 Python library to manage single and multi fasta sequences (Python 3 version). View on Gitlab
Gromacs_py Python library allowing a simplified use of the Gromacs MD simulation software. View on Github
Mol2 Python library to manage simple and multiple mol2 files. View on Gitlab
PyPDB A python parser for PDB files (Python 2 version). View on Gitlab
PyPDB3 A python parser for PDB files (Python 3 version). View on Gitlab
sdf A python parser for sdf files. View on Gitlab

C libraries

BCscore A score based on Binet-Cauchy kernel. Allows for the search of both similar and mirror conformations. Addresses two major issue of the widely used root mean square deviation (RMSD):
  • Achieves length independent statistics even for short fragments,
  • Shows better performance in the discrimination of medium range RMSD values,
  • Provides the means for large-scale mining of protein structures.
Download
QBestFit 3D superposition module, in C. Uses the quaternions. Very reliable. Download

TEF

TEF is an open-source software for decomposing protein structures into simpler yet informative units named Tightened End Fragments (or closed loops), which can be studied independently to understand protein architecture, folding, and evolution.

Lamarine M, Mornon JP, Berezovsky N, Chomilier J.
Distribution of tightened end fragments of globular proteins statistically matches that of topohydrophobic positions: towards an efficient punctuation of protein folding?
Cell Mol Life Sci. 2001 Mar;58(3):492-8.