Open-source AI tools in Structural Biology
Overview
These curated lists aim to provide an overview of the potential applications of artificial intelligence (AI) tools in addressing research questions within structural biology and pre-clinical drug discovery. It is important to note that the lists are not exhaustive, nor are they intended to represent the current state-of-the-art. Given the rapidly evolving nature of the field, each tool presents distinct advantages and limitations that should be considered in context. Only tools with publicly available codebases or accessible servers have been included. The list will be updated on a monthly basis.
Structure prediction tools
This category includes AI-driven approaches that infer the three-dimensional structures of biomolecules—such as proteins, nucleic acids, small molecules, and complexes—based on sequence or other input data. These models apply advanced machine learning techniques to capture spatial relationships, folding patterns, and atomic interactions, offering insights into molecular function and stability.
Molecular property prediction
AI-based models that estimate physicochemical, biological, and pharmacokinetic properties of molecules from structural or encoded representations. These models utilize deep learning to uncover patterns linking molecular features to behaviors such as solubility, toxicity, reactivity, affinity, and bioavailability. By automating and accelerating property prediction, they support diverse applications in structural biology and pre-clinical drug discovery.
Protein-ligand co-folding
This category includes AI models that simultaneously fold both the protein receptor and its ligand, offering an opportunity to capture orthosteric and allosteric ligands! By jointly modeling the conformational interplay between binding partners, these approaches aim to capture induced fit, allosteric modulation, and flexible binding site adaptation. Co-folding strategies enhance the realism of predicted complexes and open new avenues for understanding molecular recognition, especially in cases where static receptor assumptions fall short.
AI-guided molecular docking
This list includes computational models that use artificial intelligence to predict how small molecules bind to target proteins. Unlike classical docking methods that rely heavily on predefined scoring functions and rigid structures, AI-guided tools learn from vast datasets to model flexible interactions, improve pose accuracy, and enhance binding affinity predictions. These approaches accelerate virtual screening, support rational drug design, and offer insights into molecular recognition.
De Novo design of molecules
This list includes AI models that generate entirely new molecular structures—proteins, peptides, or small molecules—from scratch, without relying on existing templates. These generative approaches aim to optimize for specific biological functions, often targeting inhibition or binding to disease-relevant proteins. By exploring vast chemical and structural spaces, de novo design models accelerate the discovery of novel therapeutics, offering a powerful alternative to traditional trial-and-error methods in drug development and protein engineering.
Other tools
This category includes diverse AI tools that doesn’t fit to the previous groups yet are critical tools for structural biology and drug discovery research—such as identifying binding sites, predicting the impact of mutations, modeling molecular interactions, …etc. These models often complement core prediction and design workflows, offering deeper insights into molecular function and therapeutic potential.
⚠️ Disclaimer
The models listed are provided for informational purposes only. We do not endorse or guarantee the accuracy, reliability, or safety of any specific tool. Users are responsible for evaluating and using these models at their own discretion and risk.
Feedback and suggestions
For any feedback and suggestions regarding the lists, please feel free to contact us.