# Using AI to Predict Protein Shapes for Drug Discovery > This patent describes an AI-driven method for iteratively predicting protein 3D structures using a geometry neural network, then using those predictions to find and synthesize new drug molecules. - **Patent:** US 12437843 - **Original title:** Predicting protein structures using geometry neural networks that estimate similarity between predicted protein structures and actual protein structures - **Owner:** GDM Holding - **Granted:** 2025 - **Status:** Active - **Times cited:** 0 - **Field:** biotech, pharmaceutical, ai_ml, software ## What it does The patent details a method for predicting the 3D shape (structure) of a protein, which is then used to identify and create new drug molecules, called ligands. The core process involves an iterative refinement loop (Claim 1). At each step, the system maintains a "current predicted structure" and generates an "alternative predicted structure." A "geometry neural network" then processes the protein's amino acid sequence and the alternative structure's parameters to produce a "geometry score" (Claim 1). This score estimates how similar the alternative structure is to the protein's actual shape. Based on this score, the system decides whether to update its current prediction. Once a sufficiently high-quality predicted structure is determined, the method evaluates how well various "candidate ligands" (potential drugs) might interact with this predicted protein shape. It selects candidates predicted to bind effectively and then proceeds to synthesize them (Claim 1). ## What it does NOT cover - Does not cover protein structure prediction methods that do not use a geometry neural network to estimate similarity to an actual structure (Claim 1). - Does not cover methods for predicting protein structures that do not involve an iterative update process (Claim 1). - Does not cover protein structure prediction that isn't ultimately used for evaluating and selecting candidate drug ligands (Claim 1). - Does not cover methods that predict protein structures but do not include the step of synthesizing the selected ligands (Claim 1). - Does not cover protein structure prediction without considering the sequence of amino acid residues (Claim 1). ## The clever bit The core innovation is using a "geometry neural network" within an iterative refinement loop to estimate how similar a *predicted* protein structure is to its *actual* structure. This "geometry score" guides the system to improve its predictions, effectively learning to evaluate the quality of its own generated structures. ## Real-world examples 1. AlphaFold (DeepMind) 2. RoseTTAFold (University of Washington) 3. Computational drug discovery platforms 4. AI-driven pharmaceutical research ## Why it matters Accurately predicting a protein's 3D shape is a fundamental challenge in biology and crucial for designing new medicines. This patent describes an AI-driven approach to tackle this problem, which can significantly speed up drug discovery by computationally screening potential drug candidates against predicted protein targets before costly lab experiments. ## Frequently asked questions ### What does Using AI to Predict Protein Shapes for Drug Discovery cover? This patent describes an AI-driven method for iteratively predicting protein 3D structures using a geometry neural network, then using those predictions to find and synthesize new drug molecules. ### Who owns patent US 12437843? GDM Holding owns this patent, granted in 2025. ### When does this patent expire? This patent is expected to expire on September 16, 2039, when the invention enters the public domain. ### What problem does this patent solve? Accurately predicting a protein's 3D shape is a fundamental challenge in biology and crucial for designing new medicines. This patent describes an AI-driven approach to tackle this problem, which can significantly speed up drug discovery by computationally screening potential drug candidates against predicted protein targets before costly lab experiments. ### What does this patent NOT cover? Does not cover protein structure prediction methods that do not use a geometry neural network to estimate similarity to an actual structure (Claim 1). **Full plain-English explainer:** https://patentbrief.org/patent/us/12437843/predicting-protein-structures-using-geometry-neural-networks-that-estimate-simil **Original patent:** https://patents.google.com/patent/US12437843 --- _Source: PatentBrief — https://patentbrief.org. Patent facts are from public records; the plain-English explanation is PatentBrief's._ ## Related patents Semantically similar inventions in the PatentBrief corpus: - [AI Predicts 3D Protein Shapes from Genetic Code](https://patentbrief.org/patent/us/20240412809/protein-structure-prediction-from-amino-acid-sequences-using-self-attention-neur) — This patent describes how a computer uses self-attention neural networks to predict the complex three-dimensional shape of a protein based on its simple amino acid sequence, a process crucial for understanding biology and developing new medicines. - [How Computers Predict Protein Shapes Faster Using Smart Templates](https://patentbrief.org/patent/us/20210280268/protein-structure-prediction-system) — This patent describes a computer method to quickly predict the 3D shape of proteins by creating and refining "synthetic templates" from existing protein structures, reducing the heavy computational work usually needed. - [How to Predict Protein Shapes Better Using Lab Tests and Computers](https://patentbrief.org/patent/us/20210174903/enhanced-protein-structure-prediction-using-protein-homolog-discovery-and-constr) — This patent describes a method to improve predicting a protein's 3D shape by combining computer simulations with actual distance measurements from specific parts of the protein in a lab. - [Predicting How Molecules Interact with Proteins Using Two AI Streams](https://patentbrief.org/patent/us/11176462/system-and-method-for-prediction-of-protein-ligand-interactions-and-their-bioact) — This patent describes a computer system that uses two separate artificial intelligence models, one for molecules and one for proteins, to predict how they will interact and what biological effects they might have. - [Computer Model Designs Proteins by Matching Structure to Sequence](https://patentbrief.org/patent/us/12412637/embedding-based-generative-model-for-protein-design) — IBM's 2025 patent describes a computer system that designs new protein sequences by first understanding a target 3D protein shape and then generating a sequence that would fold into that shape.