Generative AI for Materials Discovery: Design Without Understanding

Engineering, Год журнала: 2024, Номер 39, С. 13 - 17

Опубликована: Июль 22, 2024

Язык: Английский

---

LMM Chemical Research with Document Retrieval

Опубликована: Авг. 13, 2024

Chemical research is more effectively progressed using Large Multimodal Models (LMMs) combined with Document Retrieval and recently published literature. The methods described here illustrate significant strides over previously tested Language Model (LLM) multi-document workflows for characterization assistance generating new reactions. Here, 3.5 Sonnet, ScholarGPT, ChatGPT 4o LMMs processed either 5 images or supplementary documents from leading 2024 journals. Each of the three models performed inference on a detailed prompt to produce response that included context attachments. In addition, were not provided which files contained answer. main findings Sonnet had an average score 9.8 images, while two judges awarded high scores (9.7, 9.4) ScholarGPT (9.5, document analysis. Judging was by human evaluator image uploads, processing evaluated Llama 3.1 405B Nemotron 4 340B LLMs correlated well improved explainability. Highlights include Sonnet's ability interpret Two-dimensional Nuclear Magnetic Resonance (2D NMR) spectrum accurately, along Judge 3.1's provide consistent formatted explanations. results shown help AI's continued revitalization established chemical field.

Язык: Английский

---

Scaffold-Lab: Critical Evaluation and Ranking of Protein Backbone Generation Methods in A Unified Framework

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Фев. 12, 2024

Abstract De novo protein design has undergone a rapid development in recent years, especially for backbone generation, which stands out as more challenging yet valuable, offering the ability to novel folds with fewer constraints. However, comprehensive delineation of its potential practical application engineering remains lacking, does standardized evaluation framework accurately assess diverse methodologies within this field. Here, we proposed Scaffold-Lab benchmark focusing on evaluating unconditional generation across metrics like designability, novelty, diversity, efficiency and structural properties. We also extrapolated our include motif-scaffolding problem, demonstrating utility these conditional models. Our findings reveal that FrameFlow RFdiffusion along Rfdiffusion GPDL showcased most outstanding performances. Furthermore, described systematic study investigate applied it task, perspective analysis methods. All data scripts will be available at https://github.com/Immortals-33/Scaffold-Lab .

Язык: Английский

---

Protein Manufacture: Protein Design Assisted by Machine Learning from Backbone to Sequence

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 337 - 346

Опубликована: Янв. 1, 2024

Язык: Английский

---

A lightweight visualization tool for protein unfolding by collision detection and elimination

Frontiers in Computer Science, Год журнала: 2024, Номер 6

Опубликована: Сен. 19, 2024

The experiments involving protein denaturation and refolding serve as the foundation for predicting three-dimensional spatial structures of proteins based on their amino acid sequences. Despite significant progress in structure engineering, exemplified by AlphaFold2 OmegaFold, there remains a gap understanding folding pathways polypeptide chains leading to final structures. We developed lightweight tool unfolding visualization called PUV whose graphics design is mainly implemented OpenGL. leverages principles from molecular biology physics, achieves rapid visual dynamics simulation chain through mechanical force atom-level collision detection elimination. After series experimental validations, we believe that this method can provide essential support investigating mechanisms pathways.

Язык: Английский

---

Progress in Multiscale Modeling of Silk Materials

Biomacromolecules, Год журнала: 2024, Номер 25(11), С. 6987 - 7014

Опубликована: Окт. 22, 2024

As a result of their hierarchical structure and biological processing, silk fibers rank among nature's most remarkable materials. The biocompatibility silk-based materials the exceptional mechanical properties certain has inspired use in numerous technical medical applications. In recent years, computational modeling clarified relationship between molecular architecture emergent demonstrated predictive power studies on novel biomaterials. Here, we review advances natural synthetic materials, from early structural silkworm cocoon to cutting-edge atomistic simulations spider nanofibrils machine learning models. We explore applications across length scales: quantum model peptides, coarse-grained dynamics proteins, finite element analysis webs. algorithmic efficiency continue advance, expect multiscale become an indispensable tool for understanding impressive developing bioinspired functional

Язык: Английский

---

Multi-purpose controllable protein generation via prompted language models

bioRxiv (Cold Spring Harbor Laboratory), Год журнала: 2024, Номер unknown

Опубликована: Ноя. 18, 2024

Deep learning is increasingly powerful for designing proteins that meet structural and functional requirements. However, most existing methods follow a conventional pipeline: first defining backbone structure then generating sequences consistent with it. This approach, which encodes all design goals indirectly through structures, restricts flexibility struggles to address multiple, complex objectives simultaneously. We present PROPEND, multi-purpose protein sequence method based on the “pre-train prompt” framework. show PROPEND’s broad utility accuracy both in silico vitro by directly controlling multiple properties prompt of backbones, blueprints, tags, their combinations. For five tested experiments, PROPEND achieved maximum recovery 105.2%, significantly outperforming classical pipeline’s 50.8%.

Язык: Английский

---

Molecular analysis and design using generative artificial intelligence via multi-agent modeling

Molecular Systems Design & Engineering, Год журнала: 2024, Номер unknown

Опубликована: Янв. 1, 2024

We report the use of a multiagent generative artificial intelligence framework, X-LoRA-Gemma large language model (LLM), to analyze, design and test molecular design. The model, inspired by biological...

Язык: Английский

---