{"id":18975735,"url":"https://github.com/babaid/ddmd","last_synced_at":"2025-09-12T11:37:42.090Z","repository":{"id":250690201,"uuid":"835179160","full_name":"babaid/DDMD","owner":"babaid","description":"A project combining a generative drug discovery model with molecular dynamics simulations.","archived":false,"fork":false,"pushed_at":"2024-08-06T16:57:56.000Z","size":108,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":1,"default_branch":"main","last_synced_at":"2024-11-08T15:21:22.652Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"gpl-3.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/babaid.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2024-07-29T10:15:29.000Z","updated_at":"2024-08-06T16:57:59.000Z","dependencies_parsed_at":"2024-08-06T19:10:23.816Z","dependency_job_id":"9db27ea8-6811-44f3-9634-24f788fa209f","html_url":"https://github.com/babaid/DDMD","commit_stats":null,"previous_names":["babaid/ddmd"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/babaid%2FDDMD","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/babaid%2FDDMD/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/babaid%2FDDMD/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/babaid%2FDDMD/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/babaid","download_url":"https://codeload.github.com/babaid/DDMD/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":232064691,"owners_count":18467521,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2024-11-08T15:20:19.568Z","updated_at":"2025-01-01T09:21:05.439Z","avatar_url":"https://github.com/babaid.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Introduction\n\nRecently deep generative models are used more and more in computer aided drug discovery. Typical models used are Diffusion Models, VAE's and different methods utilizing GNN's. In this project I will try to utilize different generative models. Here I try to focus on pocket-specific models, especially molecular linkers and fragment based models in combination with a molecular dynamics simulation. The overall goal is an improvement over a classical Docking-Simulation-Screening pipeline.\n\nA list of the models I am currently exploring:\n\n- [DeepICL](https://github.com/ACE-KAIST/DeepICL/tree/master): in theory it is the perfect model for this task as it works with interactions between a starting ligand and a pocket. In reality it doesnt seem to work with any protein, I am yet to figure out why. It also has some unclarities in design which I try to fix currently, without breaking anything.\n\n- [autofragdiff](https://github.com/ghorbanimahdi73/autofragdiff): Very promising diffusion based appraoch that also works with the binding pocket directly.\n\n\n\nSimilar to how [Cosolvent Molecular Dynamics](https://pubs.acs.org/doi/10.1021/acs.jmedchem.6b00399) allow for finding states that usually arent available at docking (e.g. binding pockets forming later on in a simulation), this proposed pipeline should also allow for similar sideffects.\n\n# Simulation Pipeline\n\nThe simulation will have following steps:\n\n1. System Setup\n2. NVT Equilibriation\n3. NPT Equilibriation\n4. Production Run for N steps\n5. Generating a new ligand based on the current topology:\n6. Repeat steps 4-5 for k times.\n\nFor now I will work with protein-ligand complex f20z used in the DeepICL example, as the binding pocket is quite well defined and yet the protein isnt too big. For simplification I deleted most of the ligand, leaving only the ring behind.\n\n# Simulation Stages\n\nFor clarity, each stage, i.e.\n    System Setup, \n    NVT, \n    NPT, \n    Production and \n    Ligand Generation \nis in a seperate script. Each step saves a state and a topology of the last simulation step and there are also reports, pdb-reports and checkpoints for safety and analysis. Each of these stages has a corresponding folder. There is also a subfolder for the ligand generation process. \n\nThis allows continuation and dynamic fixing of issues that come up during simulation, i.e. one doesnt have to start over from 0 if ligand generation fails, or can even start multiple pipelines starting from the same basis state, e.g. starting after NPT equilibriation.\n\n# Issues\n\nA possible issue is the seeting of velocities of the generated new atoms. Currently I think the easiest is to start with 0 or the thermal velocity derived from the temperature.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbabaid%2Fddmd","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fbabaid%2Fddmd","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fbabaid%2Fddmd/lists"}