惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

The GitHub Blog
The GitHub Blog
博客园 - 三生石上(FineUI控件)
V
V2EX
博客园 - 司徒正美
小众软件
小众软件
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
T
Tailwind CSS Blog
Last Week in AI
Last Week in AI
雷峰网
雷峰网
月光博客
月光博客
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
Apple Machine Learning Research
Apple Machine Learning Research
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
S
SegmentFault 最新的问题
美团技术团队
Hugging Face - Blog
Hugging Face - Blog
WordPress大学
WordPress大学
宝玉的分享
宝玉的分享
爱范儿
爱范儿
博客园 - 聂微东
量子位
J
Java Code Geeks
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
Vercel News
Vercel News

School of Computer Science News

Robotics Innovation Center Earns LEED Platinum Certification for Sustainable Construction AI4MiddleSchools Expands Nationwide Effort To Prepare Students for an AI-Powered Future Carvalho Earns NSF CAREER Award To Study Motivation and Learning Season Three of 'Does Compute' Now Available Rare Ventures Partners Rings NYSE Opening Bell Bringing Images to Life Through Touch - Robotics Institute Carnegie Mellon University Fried Receives NSF CAREER Award - Language Technologies Institute - School of Computer Science - Carnegie Mellon University PAIR Helps Students Find Their Place in AI Research Navigating the AI Era with a CMU Focus on Critical Thinking Kaess Named to Inaugural Chief of Naval Research Fellows Program Navigating the Moon Koedinger Wins Lifetime Achievement Award Carnegie Mellon Names Damion Shelton Associate VP and Executive Director of the Swartz Center for Entrepreneurship Hong Shen Discusses AI Safety at WEF Annual Meeting Erickson Earns NSF CAREER Award - Robotics Institute Carnegie Mellon University Satya Honored With Test of Time Award From Proof to Program: CMU and the Rise of AI-Driven Mathematics SCS Researchers Named to Inaugural ACM SIGSOFT Software Engineering Academy Healthcare Blind Spots: AI Models Prone To Fabricating Diagnoses - Robotics Institute Carnegie Mellon University You Can't Remove Humans From Software Engineering Designing the Future of Tech Governance AI, Single-Cell Technology Reveal How 3D Genome Differs in People With Alzheimer's Disease Tepper School of Business and School of Computer Science Partner to Launch AI for Business Executive Education Program Carnegie Mellon Researchers Lead Three DOE Genesis Mission Awards to Advance the Future of AI-Enabled Scientific Discovery Snake Robots Support Earthquake Search and Rescue in Venezuela Lindlbauer Receives NSF CAREER Award for Adaptive Extended Reality Interfaces Fredrikson Earns Test of Time Award for AI Security CMU Advances Defense Manufacturing and Military Education at Pennsylvania Defense and Innovation Summit Looking Ahead: AI Needs UI Liu Receives NSF CAREER Award
Bacteria Can Learn and Form Memories Without a Brain
Caroline She · 2026-06-05 · via School of Computer Science News
Fangwei Si

Fangwei Si was part of a team of researchers that found that bacteria can learn from past experiences.

June 5, 2026

By: Caroline Sheedy

Email
Media Inquiries

Name

Cassia Crogan

Title

University Communications & Marketing

Email

Researchers at Carnegie Mellon University have shown that bacteria can learn from past experiences, store memories across generations and adapt their behavior to changing environments all without a brain or nervous system. The research could shape how scientists think about bacterial infections and antibiotic treatment.

In a study published in PRX Life(opens in new window), researchers from CMU’s Ray and Stephanie Lane Computational Biology Department(opens in new window) and Mellon College of Science(opens in new window) tracked individual E. coli cells as nutrient conditions shifted between rich and poor environments. Instead of responding the same way every time, the bacteria adjusted their growth based on patterns they had experienced before. Cells exposed to rapidly changing conditions were able to adapt better than cells raised in more stable environments.

Josiah Kratz

Josiah Kratz

The findings suggest bacteria do more than just react to their surroundings. They appear to encode memories of past environments and use those memories to guide future behavior.

“For a long time, people assumed bacterial growth was determined only by the environment the cell is currently experiencing,” said Josiah Kratz, the first author on the paper. “What we showed is that the history of past environments matters. The cells remember those experiences, and that memory changes how they behave.”

Kratz completed most of the work as a Ph.D. student in the joint Computational Biology training program(opens in new window) between Carnegie Mellon and the University of Pittsburgh. He is now a postdoctoral fellow at the Georgia Institute of Technology.

Learning at the single-cell level

To study bacterial behavior, collaborators used a sophisticated microfluidic device known as a “mother machine,” which traps individual bacterial cells inside tiny chambers while scientists continuously observe them under a microscope.

The single-cell approach allowed researchers to rapidly switch cells between nutrient-rich and nutrient-poor environments while measuring how quickly each bacterium grew.

Fangwei Si(opens in new window), the Cooper-Siegel Assistant Professor of Physics, said this technology allowed the team to see how bacterial responses depended on current conditions, but also on how quickly conditions had fluctuated in the past. Cells exposed to fast-changing nutrient cycles adapted more quickly than cells raised in slower cycles.

“Single-cell approaches are powerful,” said Si, who specializes in biophysics, which focuses on understanding living systems in a quantitatively precise way. “As with studying animal behavior, it is more informative to learn from individual animals’ responses than from the whole population’s. Recording individual cells' life trajectories allows us to discover subtle yet fundamental details in their behavior.” 

The work builds on earlier studies showing bacteria can retain short-term memories of previous environments. 

“We showed that cells can discriminate between different environmental frequencies,” Kratz said. “To do that, they need a more complex memory than anyone had demonstrated before.”

That memory can even persist across generations. For bacteria, a generation could be tens of minutes to hours. 

A single E. coli cell may divide every 30 minutes to an hour in nutrient-rich conditions, but proteins produced during stressful conditions can be inherited by daughter and granddaughter cells. Those inherited molecules allow descendants to retain information about environments they never directly experienced.

“If a grandmother cell experienced stress and survived it, the granddaughter cell can behave differently because of that history,” Kratz said.

Implications for human health 

Traditionally, researchers assumed bacterial responses to antibiotics depended mainly on the type and concentration of drug patients are given. But if bacteria retain memories of previous stress, their responses may also depend on environmental history — bacteria that previously were exposed to starvation, high or low temperatures, or low doses of antibiotics could react differently to treatments than bacteria without that history. 

“If we want to fully understand and predict how bacteria respond to antibiotics, we may need to consider not only what they’re experiencing now, but what they experienced in the past,” Kratz said.

Researchers hope future studies will investigate whether bacteria show similar learning behaviors in response to antibiotics and other environmental stresses beyond nutrients.

The team also believes the behavior may extend beyond E. coli to many other bacterial species.

“Bacteria don’t live in static laboratory conditions in the real world,” Kratz said. “They survive in constantly changing environments like the human gut, soil or plants. Understanding how they adapt to those fluctuations is really fundamental to understanding life at the microbial level.”

In addition to the authors featured above, the research team included Huijing Wang and Shiladitya Banerjee. Banerjee was the study's corresponding author and provided the primary funding support for the research.

Connecting biology and artificial intelligence

The project brought together experiments, theory and computational modeling — a hallmark of biophysics(opens in new window), an interdisciplinary field that uses the tools of physics to understand living systems.

The researchers developed a mathematical model based on known biological processes inside bacterial cells and found that it accurately predicted how cells would respond to changing environments.

The biggest surprise came when the team analyzed the model itself.

“At an abstract level, the way the cell processes information about its environment maps onto the same type of computational architecture AI researchers use for learning problems,” Kratz said. “Biology and artificial intelligence seem to have converged on a similar strategy.”

The findings suggest that learning can emerge through chemical reactions inside a single cell, without a brain.

Under the Microscope

Image shows a "mother machine." Under a microscope, it visually resembles a tiny, translucent comb made of parallel lines, where bacteria line up single-file inside microscopic channels to divide and age while fresh growth media flows past them

A "mother machine" holds individual bacteria in tiny channels so researchers can watch them grow, divide and adapt to changing conditions over many generations.

— Related Content —