正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

Neat new system to evolve protein-protein interactions - while altering both sides of the interface! Watch for a description of the clever method, read the paper to get the deets on how these interfaces evolved & predictive models trained on screen data.

Julia Bauman

10,403 subscribers

58,527 次观看 • 2 年前 •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

5 条评论

Julia Bauman 的头像

Julia Bauman2 年前

Find the paper from @aerin_yang & coauthors here:

Soccer Bot™ 的头像

Soccer Bot™1 年前

Experience the bot completely reimagined. 🌟 A completely new design, reminders for match prediction, notifications when games are interrupted, continued or postponed. New commands like /features and /standings and new languages and much more.

ScienceStanley 的头像

ScienceStanley2 年前

Such great coverage of an amazing result. Having worked with some of these geometric learning driven molecular engineering tools, think they are going to revolutionize the whole field of compound discovery process :D

Ashton C Trotman-Grant 的头像

Ashton C Trotman-Grant2 年前

YES!👏🏾🔥

Aerin Yang 的头像

Aerin Yang2 年前

Thanks for the wonderful video, Julia!

相关视频

Introducing ESM Cambrian. Unsupervised learning can invert biology at scale to reveal the hidden structure of the natural world. We’ve scaled up compute and data to train a new generation of protein language models. ESM C defines a new state of the art for protein representation learning.

Introducing ESM Cambrian. Unsupervised learning can invert biology at scale to reveal the hidden structure of the natural world. We’ve scaled up compute and data to train a new generation of protein language models. ESM C defines a new state of the art for protein representation learning.

Alex Rives

206,727 次观看 • 1 年前

Really exciting to see ML tackle protein interactions with increasing success. A breakdown of the latest paper from Bruno Correia‘s lab:

Really exciting to see ML tackle protein interactions with increasing success. A breakdown of the latest paper from Bruno Correia‘s lab:

Julia Bauman

97,755 次观看 • 3 年前

Superior Protein for the Human Body. David delivers 28g of protein, 150 calories, and 0g of sugar, the highest protein-to-calorie ratio of any leading bar on the market. Available in 8 core indulgent flavors. Buy 4 cartons on our site, and get the 5th free.

Superior Protein for the Human Body. David delivers 28g of protein, 150 calories, and 0g of sugar, the highest protein-to-calorie ratio of any leading bar on the market. Available in 8 core indulgent flavors. Buy 4 cartons on our site, and get the 5th free.

David Protein

3,008,259 次观看 • 5 个月前

I've been building ProteoCosmos: an interactive navigator of the human protein-protein interaction network. Built on data from STRING DB, The Human Protein Atlas, AlphaFold DB, Uniprot, and PAN-GO.

I've been building ProteoCosmos: an interactive navigator of the human protein-protein interaction network. Built on data from STRING DB, The Human Protein Atlas, AlphaFold DB, Uniprot, and PAN-GO.

Brad Krajina

46,894 次观看 • 3 个月前

Get Ready for a Protein Masterclass With my PhD Advisor & Mentor Dr. Donald Layman My latest podcast is now live on all podcast platforms & on YouTube & if you want to learn a TON about protein, this is the ONE podcast you must listen to. I interview the man who had the single biggest impact on my scientific career & how I think, evaluate research, & critically appraise data. Dr Layman has published hundreds of peer reviewed scientific articles & was recently awarded the American Society for Nutrition’s Lifetime Achievement award for his contributions to nutrition science over the last 5 decades In this episode we discuss all things protein: how much should you be eating for health? To build muscle? What type of protein? Does timing matter? How does protein affect metabolism on a molecular level? Does aging affect protein needs? And so much more! You do NOT want to miss this. Learn from the man who taught me so much not just about protein and nutrition but how to read, understand, evaluate, design, & conduct research. It is my honor that my first guest is Dr. Layman.

Get Ready for a Protein Masterclass With my PhD Advisor & Mentor Dr. Donald Layman My latest podcast is now live on all podcast platforms & on YouTube & if you want to learn a TON about protein, this is the ONE podcast you must listen to. I interview the man who had the single biggest impact on my scientific career & how I think, evaluate research, & critically appraise data. Dr Layman has published hundreds of peer reviewed scientific articles & was recently awarded the American Society for Nutrition’s Lifetime Achievement award for his contributions to nutrition science over the last 5 decades In this episode we discuss all things protein: how much should you be eating for health? To build muscle? What type of protein? Does timing matter? How does protein affect metabolism on a molecular level? Does aging affect protein needs? And so much more! You do NOT want to miss this. Learn from the man who taught me so much not just about protein and nutrition but how to read, understand, evaluate, design, & conduct research. It is my honor that my first guest is Dr. Layman.

Layne Norton, PhD

17,528 次观看 • 10 个月前

As part of the release of "Both Sides (All The Sides), the new 5-LP deluxe edition of Phil Collins' fifth studio album "Both Sides" - the music videos for all four of the album's singles have been remastered in HD quality and are now available to watch on YouTube.

As part of the release of "Both Sides (All The Sides), the new 5-LP deluxe edition of Phil Collins' fifth studio album "Both Sides" - the music videos for all four of the album's singles have been remastered in HD quality and are now available to watch on YouTube.

Phil Collins

16,726 次观看 • 1 年前

How to lose fat and gain muscle at the same time: – Keep calories 10% above maintenance, roughly 200 to 300 calories over. The surplus should come primarily from quality protein. – Protein is the key variable: aim for 1 to 1.5 grams per pound of body weight. Studies show that adding 80 to 100 grams of protein on top of habitual intake leads to fat loss, not fat gain. – Distribution of protein throughout the day matters less than hitting the total daily target. Alan Aragon on the Andrew D. Huberman, Ph.D. podcast

How to lose fat and gain muscle at the same time: – Keep calories 10% above maintenance, roughly 200 to 300 calories over. The surplus should come primarily from quality protein. – Protein is the key variable: aim for 1 to 1.5 grams per pound of body weight. Studies show that adding 80 to 100 grams of protein on top of habitual intake leads to fat loss, not fat gain. – Distribution of protein throughout the day matters less than hitting the total daily target. Alan Aragon on the Andrew D. Huberman, Ph.D. podcast

Huberman Lab Clips

261,091 次观看 • 27 天前

Candace Owens plays a clip of Ana Kasparian on her show & explains how American citizens on both sides of the political aisle get smeared ruthlessly in the press for challenging lobbyist narratives

Candace Owens plays a clip of Ana Kasparian on her show & explains how American citizens on both sides of the political aisle get smeared ruthlessly in the press for challenging lobbyist narratives

Irlandarra

151,287 次观看 • 6 个月前

Candace Owens plays a clip of Ana Kasparian on her show & explains how American citizens on both sides of the political aisle get smeared ruthlessly in the press for challenging lobbyist narratives.

Candace Owens plays a clip of Ana Kasparian on her show & explains how American citizens on both sides of the political aisle get smeared ruthlessly in the press for challenging lobbyist narratives.

Sulaiman Ahmed

62,426 次观看 • 6 个月前

How to extract data from papers for literature review in seconds? 1. Go to 2. Upload 2-3 papers you already know are relevant 3. Start with clicking on one paper 4. ResearchCollab do the following for the paper ✦ Provide a short overview of the paper ✦ Identify weaknesses in the paper ✦ Share papers that contrast the existing paper ✦ Extract meta data about the paper ✦ Evaluate each part (e.g., methodology) of the paper ✦ Provide AI chat to extract any other data 5. Now click on the Related Papers button at top right 6. You will get papers related to the seed paper 7. You can click on each paper and read its abstract 8. If you find it relevant, just add it to your list of papers 9. Do the same for the other 2 seed papers 10. This way you will collect all relevant papers 11. And extract data from those papers After this, analyse the data and report the findings. Try ResearchCollab today:

How to extract data from papers for literature review in seconds? 1. Go to 2. Upload 2-3 papers you already know are relevant 3. Start with clicking on one paper 4. ResearchCollab do the following for the paper ✦ Provide a short overview of the paper ✦ Identify weaknesses in the paper ✦ Share papers that contrast the existing paper ✦ Extract meta data about the paper ✦ Evaluate each part (e.g., methodology) of the paper ✦ Provide AI chat to extract any other data 5. Now click on the Related Papers button at top right 6. You will get papers related to the seed paper 7. You can click on each paper and read its abstract 8. If you find it relevant, just add it to your list of papers 9. Do the same for the other 2 seed papers 10. This way you will collect all relevant papers 11. And extract data from those papers After this, analyse the data and report the findings. Try ResearchCollab today:

Faheem Ullah

26,494 次观看 • 4 个月前

🧬A new imaging technique developed by the Liu Lab uses a novel DNA barcode system to track hundreds of RNA & protein molecules in single cells within thick biological samples, providing a full picture of how these structures are organized inside tissues.

🧬A new imaging technique developed by the Liu Lab uses a novel DNA barcode system to track hundreds of RNA & protein molecules in single cells within thick biological samples, providing a full picture of how these structures are organized inside tissues.

HHMI | Janelia

79,436 次观看 • 1 年前

#NewPaper The first microscope, invented in the 16th century, was designed to unlock the secrets of the microscopic world. Today, as many fields become increasingly data-driven, there is a pressing need for new types of microscopes---tools that help us zoom in, explore, and understand complex data. We call these tools "algorithmic microscopes." Introducing the Vendiscope: The first algorithmic microscope for data collections. 🔬 The Vendiscope maximizes the probability-weighted Vendi Score of a dataset to assign a weight to each element in the collection. This weight represents a data point's contribution to the overall diversity of the collection. These weights enable high-resolution data analysis at scale. We use them to zoom in on datasets across three domains: biology, materials science, & AI. 🧬 Biology: We used the Vendiscope on the protein universe, which contains nearly 250 million proteins. We found that nearly 200 million of the proteins are near-duplicates of each other and that AlphaFold fails on proteins that contribute most to the diversity of the protein universe. (See GIF below). 🪜 Materials Science: We used the Vendiscope on the Materials Project database, which contains 170K materials as of today. We found that 85% of crystals with formation energy data are near-duplicates of each other and that ML models for materials property prediction struggle with materials that contribute most to diversity. 🤖 Artificial Intelligence: We applied the Vendiscope to CIFAR-10, a benchmark dataset containing 50K images. We found duplicates. We applied the Vendiscope to analyze state-of-the-art generative models trained on this dataset. We found the best generative models memorize training data, as is known in the AI literature. However, we can do more with the Vendiscope and characterize the type of samples that get memorized. We found that data points contributing least to diversity are more prone to memorization by these generative models. 🧠 "Our findings demonstrate that the Vendiscope can serve as a powerful tool for data-driven science, providing a systematic and scalable way to identify duplicates and outliers, as well as pinpointing samples prone to memorization and those that models may struggle to predict---even before training." 💫 "The Vendiscope provides a unified framework for analyzing complex data at scale. Researchers, engineers, and data auditors can use the Vendiscope to audit datasets, identify potential biases, and refine data collection practices. For AI ethicists, the Vendiscope offers a critical lens to understand how models interact with data, particularly in the context of bias, memorization, and data fairness, enabling better mitigation strategies to prevent undesirable outcomes in AI deployment. For scientists, the Vendiscope represents a new companion in the discovery process." #VendiScoring #AlgorithmicMicroscopy Link to paper: Authors: Amey Pasarkar (Amey Pasarkar) and Adji Bousso Dieng (@adjiboussodieng)

#NewPaper The first microscope, invented in the 16th century, was designed to unlock the secrets of the microscopic world. Today, as many fields become increasingly data-driven, there is a pressing need for new types of microscopes---tools that help us zoom in, explore, and understand complex data. We call these tools "algorithmic microscopes." Introducing the Vendiscope: The first algorithmic microscope for data collections. 🔬 The Vendiscope maximizes the probability-weighted Vendi Score of a dataset to assign a weight to each element in the collection. This weight represents a data point's contribution to the overall diversity of the collection. These weights enable high-resolution data analysis at scale. We use them to zoom in on datasets across three domains: biology, materials science, & AI. 🧬 Biology: We used the Vendiscope on the protein universe, which contains nearly 250 million proteins. We found that nearly 200 million of the proteins are near-duplicates of each other and that AlphaFold fails on proteins that contribute most to the diversity of the protein universe. (See GIF below). 🪜 Materials Science: We used the Vendiscope on the Materials Project database, which contains 170K materials as of today. We found that 85% of crystals with formation energy data are near-duplicates of each other and that ML models for materials property prediction struggle with materials that contribute most to diversity. 🤖 Artificial Intelligence: We applied the Vendiscope to CIFAR-10, a benchmark dataset containing 50K images. We found duplicates. We applied the Vendiscope to analyze state-of-the-art generative models trained on this dataset. We found the best generative models memorize training data, as is known in the AI literature. However, we can do more with the Vendiscope and characterize the type of samples that get memorized. We found that data points contributing least to diversity are more prone to memorization by these generative models. 🧠 "Our findings demonstrate that the Vendiscope can serve as a powerful tool for data-driven science, providing a systematic and scalable way to identify duplicates and outliers, as well as pinpointing samples prone to memorization and those that models may struggle to predict---even before training." 💫 "The Vendiscope provides a unified framework for analyzing complex data at scale. Researchers, engineers, and data auditors can use the Vendiscope to audit datasets, identify potential biases, and refine data collection practices. For AI ethicists, the Vendiscope offers a critical lens to understand how models interact with data, particularly in the context of bias, memorization, and data fairness, enabling better mitigation strategies to prevent undesirable outcomes in AI deployment. For scientists, the Vendiscope represents a new companion in the discovery process." #VendiScoring #AlgorithmicMicroscopy Link to paper: Authors: Amey Pasarkar (Amey Pasarkar) and Adji Bousso Dieng (@adjiboussodieng)

Vertaix® (AI & Science)

34,762 次观看 • 1 年前

Excited to announce our new product: AI Protein Design Copilot Tamarind Bio has been the leading provider to use the top AI tools in computational protein engineering. Today we announce a completely new way to interface with computational molecular design. Available now!

Excited to announce our new product: AI Protein Design Copilot Tamarind Bio has been the leading provider to use the top AI tools in computational protein engineering. Today we announce a completely new way to interface with computational molecular design. Available now!

Deniz Kavi

33,356 次观看 • 11 个月前

📣Just released a paper on State repression of #Environmental #Protest and #CivilDisobedience. The repression faced by #EnvironmentalDefenders is a threat to #HumanRights & #Democracy. Read the thread below for my calls to action and read the full paper 👇

📣Just released a paper on State repression of #Environmental #Protest and #CivilDisobedience. The repression faced by #EnvironmentalDefenders is a threat to #HumanRights & #Democracy. Read the thread below for my calls to action and read the full paper 👇

Michel Forst

69,973 次观看 • 2 年前

I now have the world’s largest collection of these old films, showing how basic processes worked. I’ve made great progress training AI on 100s of these thus far—I have 100s more to go, but it’s quite astonishing the details of remembrance. This is high protein training data.

I now have the world’s largest collection of these old films, showing how basic processes worked. I’ve made great progress training AI on 100s of these thus far—I have 100s more to go, but it’s quite astonishing the details of remembrance. This is high protein training data.

Brian Roemmele

385,060 次观看 • 6 个月前

$.Michael Nielsen on AlphaFold - the AI bit is a small part of the total investment. "A massive fraction of the success there is the Protein Data Bank. It's X-ray diffraction, NMR, cryo-EM, and the several billion dollars that was spent obtaining 180,000 protein structures."$

.Michael Nielsen on AlphaFold - the AI bit is a small part of the total investment. "A massive fraction of the success there is the Protein Data Bank. It's X-ray diffraction, NMR, cryo-EM, and the several billion dollars that was spent obtaining 180,000 protein structures."

Dwarkesh Patel

110,495 次观看 • 2 个月前

AI is trained on the work, data, and creativity of millions of people. We the people deserve a say in how it shapes our economy & society. Watch the full interview on @meidastouch's youtube channel.

AI is trained on the work, data, and creativity of millions of people. We the people deserve a say in how it shapes our economy & society. Watch the full interview on @meidastouch's youtube channel.

Saikat Chakrabarti for Congress

11,207 次观看 • 3 个月前

YC F24's Reticular (YC F24) makes protein AI models controllable and interpretable to help steer protein design with limited biological data, reducing costly validation cycles. Congrats on the launch, Nithin Parsan and John Yang!

YC F24's Reticular (YC F24) makes protein AI models controllable and interpretable to help steer protein design with limited biological data, reducing costly validation cycles. Congrats on the launch, Nithin Parsan and John Yang!

Y Combinator

29,575 次观看 • 1 年前

The debunk of the 2 Marshmallow Test has been debunked & most people don’t cite these studies correctly. What the actual experiment was & data ultimately show. The takeaway: ability to delay gratification is on a continuum & yes it is predictive of (certain aspects of) success.

The debunk of the 2 Marshmallow Test has been debunked & most people don’t cite these studies correctly. What the actual experiment was & data ultimately show. The takeaway: ability to delay gratification is on a continuum & yes it is predictive of (certain aspects of) success.

Andrew D. Huberman, Ph.D.

179,173 次观看 • 1 个月前