Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: biotech/medical

Viewing Neural Networks Through a Statistical-Physics Lens

Statistical physics is shedding light on how network architecture and data structure shape the effectiveness of neural-network learning.

Machine-learning technologies have profoundly reshaped many technical fields, with sweeping applications in medical diagnosis, customer service, drug discovery, and beyond. Central to this transformation are neural networks (NNs), models that learn patterns from data by combining many simple computational units, or neurons, linked by weighted connections. Acting collectively, these neurons can process data to learn complex input–output relationships. Despite their practical success, the fundamental mechanisms by which NNs learn remain poorly understood at a theoretical level. Statistical physics offers a promising framework for exploring central questions in machine-learning theory, potentially clarifying how learning depends on the layout of the network—the NN architecture—and on statistics of the data—the data structure (Fig. 1).

Three recent papers in a special Physical Review E collection (See Collection: Statistical Physics Meets Machine Learning — Machine Learning Meets Statistical Physics) provide significant insights into these questions. Francesca Mignacco of City University of New York and Princeton University and Francesco Mori of the University of Oxford in the UK derived analytical results on the optimal fraction of neurons that should be active at a given time [1]. Abdulkadir Canatar and SueYeon Chung of the Flatiron Institute in New York and New York University investigated the influence of the precision with which a network is “trained” on the amount of data the NN can reliably decode [2]. Francesco Cagnetta at the International School for Advanced Studies in Italy and colleagues showed that NNs whose structure mirrors that of the data learn faster [3].

Why do microbes team up? A new model explains nutrient sharing in fluctuating environments

Depending on others for something you need may feel like a risky proposition—and perhaps a human one. It is actually a survival strategy found in the microbial world, and far more frequently than one might expect. Discovering why is key to understanding how microbes form stable communities across medical, industrial, and ecological settings.

A new study by bioengineering professor Sergei Maslov (CAIM co-leader), computational scientist Ashish George, and biology professor Tong Wang explores why interdependence can be such a winning move for microbial communities. Their work, published in Cell Systems, demonstrated that a mathematical model of how bacteria produce and share resources accurately predicted the outcome of experiments with living E. coli strains.

The researchers’ collaboration began during their time as colleagues at the Carl R. Woese Institute for Genomic Biology at the University of Illinois Urbana-Champaign. George continued the collaboration in his position at the Broad Institute; Wang, in his appointment at Purdue University. Maslov, who led the study, remains at Illinois and is an affiliate member of the National Institute for Theory and Mathematics in Biology.

The AI Tsunami is Here & Society Isn’t Ready | Dario Amodei x Nikhil Kamath | People by WTF

I sat down with Dario Amodei in Bangalore. He built Claude, but he started as a biologist looking for a tool to cure disease. Today, he’s at the helm of an AI revolution that he compares to a tsunami society is actively ignoring. We got into the heavy stuff: why Anthropic secretly withheld a working model before ChatGPT existed, whether AI is on the verge of consciousness, and if outsourcing our thinking is going to make humans measurably stupider. Dario makes the case that coding is a dying skill, critical thinking is our last real edge, and the absurd concentration of power in AI right now is a massive problem, even though he’s one of the people holding it.

00:00 Introduction.
06:13 Scaling laws explained simply.
13:27 Trust, humility, and corporate motives.
22:44 Using Claude personally, AI knowing you.
31:03 Rich people criticizing their own system.
37:05 India’s role and IT partnerships.
44:15 Will AI surpass humans at everything.
50:17 Career advice for young Indians.
56:38 Open source vs closed AI models.
1:02:40 Biotech as the next big bet.

#NikhilKamath Co-founder of Zerodha and Gruhas.
Host of ‘WTF is’ & ‘People By WTF’ Podcast.
Twitter: https://twitter.com/nikhilkamathcio/
Instagram: / nikhilkamathcio.
LinkedIn: https://www.linkedin.com/in/nikhilkam… / nikhilkamathcio #Darioamodei LinkedIN– / dario-amodei X — https://twitter.com/DarioAmodei Instagram — / dario.amodei Watch ‘WTF is’ Podcast on Spotify https://tinyurl.com/4nsm4ezn Watch ‘People by WTF’ Podcast on Spotify https://tinyurl.com/yme92c59 Watch ‘WTF Online’ on Spotify https://tinyurl.com/4tjua4th #WTFiswithnikhilkamath #PeopleByWTF #WTFOnline.
Facebook: / nikhilkamathcio.

#Darioamodei.
LinkedIN-/ dario-amodei.
X — https://twitter.com/DarioAmodei.
Instagram — / dario.amodei.

Watch ‘WTF is’ Podcast on Spotify.
https://tinyurl.com/4nsm4ezn.

Watch ‘People by WTF’ Podcast on Spotify.

Continue reading “The AI Tsunami is Here & Society Isn’t Ready | Dario Amodei x Nikhil Kamath | People by WTF” | >

Chemists thought phosphorus had shown all its cards—until it surprised them with a new move

A discovery by UCLA organic chemists may one day put catalytic converter thieves out of business. In new research, they’ve used abundant, inexpensive phosphorus as a catalyst in chemical reactions that usually require precious metals like platinum, one of the metals targeted in theft of the automotive components that convert chemicals in vehicle exhaust into less harmful forms.

This advance, however, will likely be more useful in the pharmaceutical industry and could one day help bring down the price of some drugs.

Common anti-seizure drug prevents Alzheimer’s plaques from forming

At the heart of the new discovery is amyloid precursor protein (APP), a protein that plays important roles in brain development and synaptic formation. Abnormal processing of APP can lead to the production of amyloid‑beta peptides, which play a central role in the development of Alzheimer’s disease. The scientists found that how APP is trafficked also controls whether a neuron forms amyloid-beta 42.

During the synaptic vesicle cycle — a fundamental process that underlies every thought, movement, memory or sensation — levetiracetam binds to a protein called SV2A. This interaction slows down a step in which neurons recycle synaptic vesicle components from the cell’s surface. By pausing this recycling process, the drug enables APP to remain on the cell’s surface longer, diverting it away from the pathway that produces toxic amyloid‑beta 42 proteins.

“In our 30s, 40s and 50s, our brains are generally able to steer proteins away from harmful pathways,” the author said. “As we age, that protective ability gradually weakens. This is not a statement of disease; this is just a part of aging. But in brains developing Alzheimer’s, too many neurons go astray, and that’s when you get amyloid-beta 42 production. And then it’s tau (or ‘tangles’), and then it’s dead cells, then dementia, then neuroinflammation — and then it’s too late.”

To effectively prevent Alzheimer’s symptoms, high-risk individuals would need to begin taking levetiracetam “very, very early,” the author said, possibly up to 20 years before the new FDA-approved Alzheimer’s disease test would even capture mildly elevated levels of amyloid-beta 42.

“You couldn’t take this when you already have dementia because the brain has already undergone a number of irreversible changes and a lot of cell death,” the author said.

Leveraging its status as an FDA-approved and widely used drug, the team mined existing human clinical data to investigate whether Alzheimer’s patients who took levetiracetam experienced slowed cognitive decline. They obtained clinical data from the National Alzheimer’s Coordinating Center and conducted a correlative analysis, finding that Alzheimer’s patients who took levetiracetam were associated with a significant delay from the diagnosis of cognitive decline to death compared to those taking lorazepam or no/other anti-epileptic drugs. ScienceMission sciencenewshighlights.

Continue reading “Common anti-seizure drug prevents Alzheimer’s plaques from forming” | >

This Research Article uncovers an unexpected function of iron regulatory protein 1 in metabolic regulation

Kostas Pantopoulos & team find mice that lack IRP1 have altered energy metabolism and are protected against metabolic syndrome pathologies:

The figure shows liver in Irp1-/- mice fed a high-fat diet have reduced fat content; stained with oil red O. MetabolicSyndrome.

1Lady Davis Institute for Medical Research, Jewish General Hospital and Department of Medicine, McGill University, Montreal, Quebec, Canada.

2Department of Biochemistry, Microbiology & Immunology, University of Ottawa, Ottawa, Ontario, Canada.

3Institute for Research in Immunology and Cancer, University of Montreal, Montreal, Quebec, Canada.

New lab technique can reverse chemical process linked with Alzheimer’s disease

An Oregon State University scientist and a team of undergraduate students have uncovered real-time insights into a chemical process linked with Alzheimer’s disease, paving the way toward better drug designs. The researchers used a molecule measuring technique to observe in a laboratory setting how certain metals can promote the protein clumping that leads to the blocked neural pathways associated with Alzheimer’s. Led by Marilyn Rampersad Mackiewicz, associate professor of chemistry in the OSU College of Science, the research team also watched molecules known as chelators disrupt or reverse the clumping. The findings are published in ACS Omega.

Alzheimer’s disease is the most common form of dementia, a chronic condition of impaired cognitive function that affects large numbers of older adults and their loved ones. According to the Centers for Disease Control and Prevention, Alzheimer’s is the sixth-leading cause of death for people age 65 and older.

In Alzheimer’s patients, aggregations of amyloid-beta proteins interrupt brain cells’ ability to communicate with each other. The brain needs certain metals to work properly, but problems arise when the metals are present in unbalanced quantities.

Abstract: Birt-Hogg-Dubé syndrome is largely characterized by pulmonary cysts, but disease models for this lung phenotype are lacking

Here, Elizabeth P. Henske report mTORC1 hyperactivation in cystic lungs from patients and validate its role in lung cyst formation using a novel pre-clinical model:

The figure shows lung mesenchymal and epithelial cell differentiation.

1Division of Pulmonary, Critical Care and Sleep Medicine; Department of Internal Medicine; University of Cincinnati College of Medicine, Cincinnati, Ohio, USA.

2The Saban Research Institute, Children’s Hospital Los Angeles, Keck School of Medicine, University of Southern California, Los Angeles, California, USA.

3Division of Pulmonary and Critical Care Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, USA.

/* */