Complexity Thoughts: Issue #15
Unraveling complexity: building knowledge, one paper at a time
If you find Complexity Thoughts, click on the Like button, leave a comment, repost on Substack or share this post. It is the only feedback I can have for this free service.
The frequency and quality of this newsletter relies on social interactions. Thank you!
Network & Complex Systems foundations
Imagine trying to understand how gases or materials like magnets interact with each other, not just with nearby particles, but also with ones far away. The authors have found a groundbreaking way to do this by using Monte Carlo simulations, which are like rolling the dice to create random states of a system. The groundbreaking news is they've developed a super-fast method to do this. In the past, it would take centuries, but now it's done in days. They're particularly interested in studying systems with long-range interactions, where particles can affect each other even if they're distant. This is crucial because it can helps us explore processes in various fields, from droplets forming on a window during a hot shower to understanding how cells work.
This new approach opens up countless possibilities for research and practical applications, showing the power of computer simulations in modern physics, alongside experiments and analytical methods. It's like having a high-speed microscope to explore the secrets of the universe at a whole new level.
Origin of life
Self-organization of primitive metabolic cycles due to non-reciprocal interactions
One of the greatest mysteries concerning the origin of life is how it has emerged so quickly after the formation of the earth. In particular, it is not understood how metabolic cycles, which power the non-equilibrium activity of cells, have come into existence in the first instances. While it is generally expected that non-equilibrium conditions would have been necessary for the formation of primitive metabolic structures, the focus has so far been on externally imposed non-equilibrium conditions, such as temperature or proton gradients. Here, the authors propose an alternative paradigm in which naturally occurring non-reciprocal interactions between catalysts that can partner together in a cyclic reaction lead to their recruitment into self-organized functional structures. They uncover different classes of self-organized cycles that form through exponentially rapid coarsening processes, depending on the parity of the cycle and the nature of the interaction motifs, which are all generic but have readily tuneable features.
Synthetic/Systems/Molecular Biology
Engineered bacteria detect tumor DNA
“Synthetic biology has developed sophisticated cellular biosensors to detect and respond to human disease. However, biosensors have not yet been engineered to detect specific extracellular DNA sequences and mutations. Here, we engineered naturally competent Acinetobacter baylyi to detect donor DNA from the genomes of colorectal cancer (CRC) cells, organoids, and tumors. We characterized the functionality of the biosensors in vitro with coculture assays and then validated them in vivo with sensor bacteria delivered to mice harboring colorectal tumors. We observed horizontal gene transfer from the tumor to the sensor bacteria in our mouse model of CRC. This cellular assay for targeted, CRISPR-discriminated horizontal gene transfer (CATCH) enables the biodetection of specific cell-free DNA.”
Fanzor is a eukaryotic programmable RNA-guided endonuclease
Scientists have discovered a fascinating system that uses RNA to guide molecular "scissors" (called endonucleases) that can cut DNA. This cutting action is crucial because it helps cells make precise changes to their genetic instructions, a bit like editing a manual. Endonucleases are like the cell's tools for genetic surgery. They can be used to fix DNA errors or make specific changes to genes.
In this paper, authors have found these RNA-guided systems in both simple and complex organisms, from single-celled bacteria to humans. For instance, in bacteria, there's a system known as CRISPR–Cas, which acts as a defense mechanism against invaders by cutting their DNA. It's like providing bacteria with their simple-yet-complex own immune system.
The finding means that even in higher life forms there's a tool similar to the molecular scissors found in bacteria.
A breakthrough discovery: the cell nucleus, previously thought to be inactive, becomes highly active during crises like DNA damage. This fact challenges our traditional view of the nucleus as a passive, non-metabolic part of the cell. Instead, it suggests that the nucleus can be quite active during emergencies. When cells face threats, such as widespread DNA damage, they deploy mitochondrial enzymes to the nucleus to address the issue. This newfound metabolic activity in the nucleus has significant implications for cancer research. Understanding these processes could aid in overcoming drug resistance and designing more effective cancer treatments. The authors also identified an enzyme, PRDX1, that plays a critical role in mitigating the effects of DNA damage. This study highlights the dynamic nature of cellular responses to stress and the potential for innovative cancer therapies.
Science and Artificial Intelligence
Scientific discovery in the age of artificial intelligence
“Artificial intelligence (AI) is being increasingly integrated into scientific discovery to augment and accelerate research, helping scientists to generate hypotheses, design experiments, collect and interpret large datasets, and gain insights that might not have been possible using traditional scientific methods alone. Here we examine breakthroughs over the past decade that include self-supervised learning, which allows models to be trained on vast amounts of unlabelled data, and geometric deep learning, which leverages knowledge about the structure of scientific data to enhance model accuracy and efficiency. Generative AI methods can create designs, such as small-molecule drugs and proteins, by analysing diverse data modalities, including images and sequences. We discuss how these methods can help scientists throughout the scientific process and the central issues that remain despite such advances. Both developers and users of AI tools need a better understanding of when such approaches need improvement, and challenges posed by poor data quality and stewardship remain. These issues cut across scientific disciplines and require developing foundational algorithmic approaches that can contribute to scientific understanding or acquire it autonomously, making them critical areas of focus for AI innovation.”
AI tools are designing entirely new proteins that could transform medicine
Scientists are harnessing the power of artificial intelligence to create proteins that nature has never produced in billions of years of evolution. These custom-designed proteins hold enormous potential for applications (vaccines, therapeutics, biomaterials, …).
These AI tools work on a principle similar to how AI generates realistic images. They take random sequences of amino acids and refine them through denoising, to produce realistic protein structures. What's remarkable is that researchers can guide these tools to create proteins with specific features or functions. For instance, they can design proteins to bind tightly to other biomolecules (including those involved in diseases like cancer and autoimmune conditions?)
From the perspective of AI(especially the generative model).
We will be happy if there is a domain with a lot of accessible data(like language, pictures, proteins).
We will be happier if we manage to find a good way to represent the data(like a vector for a word, a matrix for a picture).
We will be even happier if we can label each single data(this is a good answer in a dialog, this is a smiling/crying face).
If all the conditions were satisfied, conditional generative models can be used to understand these data, including finding the distribution and generating some entirely new data under some human desire(condition).
I'm happy to see that over years protein domain has overcome many difficulties to reach todays result, in my memory(if I remember correctly) some years ago they were struggling to find a good representation of a protein structure.(I roughly looked at the paper, in this paper it seems that they proposed a new method to represent a peotein, it may be helpful to other researchers in this domain).