310.ai

Our copilot is live!

🚀 Dive into the groundbreaking world of AI-generated protein sequences! Our latest breakthrough features a kinase designed by the MPM4 AI text2protein model. 🔬 Why Kinases? Kinases are pivotal in cellular signaling and are key targets in treating diseases like cancer and diabetes. Our AI-driven approach is not just enhancing how we understand these proteins; it's revolutionizing how we use them. Explore: https://lnkd.in/gdAvQhxS Visualize: https://lnkd.in/gz8jy34n Design: https://310.ai/copilot/

No code needed. With 1 line of english text, modify your PDB visualization. https://lnkd.in/gz8jy34n Try it for yourself in https://310.ai/copilot

The National Academies of Sciences Engineering and Medicine (NASEM) is hosting a meeting Tue Aug 13-14 on ⭐ Assessing and Navigating Biosecurity Concerns and Benefits of Artificial Intelligence Use in the Life Sciences ⭐ Register (online available) here: https://lnkd.in/gyeAW6Zj

🧬🧺⚕️ Detergents, stroke therapy, cell cultures, oh my! Meet peptidases, the 'protein scissors' making it all possible! Dive into our latest video and see how our AI transforms text into novel proteins with precision. Explore the future with molecule programming AI technology. Repo: https://310.ai/mpm/repo Copilot: https://lnkd.in/gfqzgX3E

Find a de novo ✨ AI generated protein and take it for a test drive. Novel MPM4 text2protein repo samples can now be pulled into copilot for testing and further programming. What will you be able to create? Check out the repo here: https://310.ai/mpm/repo Check out copilot here: https://lnkd.in/gEsPvtqB 

🤖 Unveiling the Future: Text2Protein Modeling 👨💻 The team at 310.ai just released there Text to Protein modeling technology. This innovation merges AI with molecular biology, unlocking numerous possibilities. Let's take a look at how this technology is changing the industry 🔽 Applications of Text to Protein Modeling Drug Discovery 💊 Accelerates the identification of drug-binding sites, aiding in the development of effective drugs with fewer side effects. Personalized Medicine 🩺 Tailors treatments based on individual genetic profiles, improving efficacy and reducing adverse reactions. Disease Understanding 🦠 Visualizes protein structures involved in diseases, leading to better insights and new therapeutic targets. Protein Engineering 🏗️ Designs new proteins with desired functions for applications in industrial biotechnology, such as biofuel production and novel materials. Educational Tools 📚 Enhances learning by allowing students to visualize 3D structures of proteins from textual descriptions. Benefits of Text to Protein Modeling Speed and Efficiency ⚡ Streamlines protein modeling, allowing rapid generation and analysis. Accessibility 🌐 Makes protein modeling accessible without extensive computational resources. Innovation 🌟 Fosters new discoveries and applications in science and medicine. Credit to Kooshiar Azimian, Kathy Y. Wei, Ph.D. and their team on the amazing accomplishment! #AI #Innovation #Biotech #Pharma

310 AI has just launched 𝐌𝐏𝐌4 (Molecular Programming Model v4), a new generative natural language text-to-protein sequence foundation model. This is another HUGE step towards making Biology more programmable. In protein science, there are two long-standing and widely tackled problems: ◾ the protein folding (structure prediction from sequence) problem, and ◾ the inverse folding (sequence design for given structure) problem. With the emergence of DeepMind’s AlphaFold2 in 2020, the protein folding (sequence → structure) problem was considered (largely) “solved”, with more and more advances succeeding AF2. Following AF2, there have been new DL-based models (ProteinMPNN/LigandMPNN, ESM-IF, etc.) with excellent performance towards the inverse folding (structure → sequence) problem. Today, 310.ai launched MPM, a model developed to tackle a 𝐭𝐡𝐢𝐫𝐝 𝐩𝐫𝐨𝐛𝐥𝐞𝐦 which, until recently, hasn’t received much attention. And that is the “Programming Problem” (designing novel proteins from specified functions). Recent models like Chroma and ESM3 have made significant strides in advancing the programmability direction. MPM4 takes this multiple steps further! Here are a few key things you should know about MPM: ◾ Given a natural language description as a prompt, MPM can generate protein sequences that fit the described functions and/or properties. The prompt can be as short as a single-line sentence. ◾ MPM is nicely integrated with other protein modeling tools for validation of MPM designs via structure modeling and function prediction. ◾ These integrations provide myriads of functional, structural (folding), and sequence evaluation/prediction scores for generated sequences. ◾ Structural evaluation score (pLDDT) is generated by folding the sequence with ESMFold. ◾ Function prediction is done with ProtNLM and (text-based) relevance to input prompt is evaluated with GPT ◾ Sequence novelty scores are generated using DIAMOND. ◾ Structurally related structures and extent are evaluated using Foldseek search from databases. ◾ 310 AI released an MPM4 repository of 1000+ novel AI-generated, high quality, and carefully examined protein sequences covering a wide range of functions (link below). Great work by Kooshiar Azimian, Kathy Y. Wei, Ph.D., and the 310.ai team! While this is certainly a groundbreaking advancement for protein design/engineering and biomolecular programming,   the next milestone for MPM is perhaps wet lab structural and functional evaluation of some of the high quality designs. White paper: https://lnkd.in/gp2gcHm8 MPM4 repo: https://310.ai/mpm/repo - - - I’ve been exploring the MPM4 repo and intend to share some interesting findings later. These will also be featured in the next issue (this weekend) of the BioMoDes newsletter (subscribe here https://lnkd.in/gzPR38VP) BTW, I made a post about the 310.ai copilot months ago (Link in the comment). Really cool stuff.

🚀 Exciting AI-Driven Protein Design! Recently, I evaluated the innovative Text2Protein AI model from 310.ai (kudos Kathy Y. Wei, Ph.D., and Kooshiar Azimian, whitepaper: http://bit.ly/310paper), which generates new protein sequences from simple prompts. Protein Design 🔍 Features: Histidine kinase, PAS-associated domain Prompt: The protein features a histidine kinase domain along with a PAS-associated domain located at the C-terminal. Result: https://lnkd.in/ejxDM5qe The protein structure can be downloaded in PDB format for further analysis. Metrics: pLDDT: 93.00 (Very High Confidence) Predicted Local Distance Difference Test score, indicating confidence in the predicted protein structure. This per-residue analysis can be easily mapped onto a structure to display the confidence of local features like α-helices. A score above 70 generally suggests high confidence. A score of ~93 indicates very high confidence in the structural prediction. Pident (Seqdif): 39.90% (Novelty) Percentage Identity of the sequence alignment, i.e. the similarity between designed protein and reference protein in the NCBI database. Lower Pident suggests greater novelty.   A quick analysis of the sequences for functional insights revealed that the newly generated sequences closely resemble Response Regulator Proteins. These proteins, featuring N-terminal receiver and PAS sensory domains, are crucial in bacterial signal transduction. This design suggests potential roles in similar pathways. Evolutionary Context: 🌍 Conserved Across Bacteria: Indicates a fundamental role in cellular processes, crucial for bacterial adaptation, metabolism, and antibiotic resistance. Conclusion: The Text2Protein AI model from 310.ai showcases impressive potential in designing novel proteins with significant structural and functional relevance. This could be a game-changer in developing new therapeutic strategies! 🧬 #AI #ProteinDesign #MachineLearning

First text2protein AI model, compressing billions of years of life. 800+ novel, functional and foldable proteins are discovered by researchers within the published repo. Whitepaper and repo https://bit.ly/310paper