Nearly 1,000 AI agents searched nature’s enzymes—and uncovered a mysterious DNA system with CRISPR-like clues. Could AI help decode life, design new proteins, and reshape medicine? We explore the science, the risks, and the big ethical questions. The discovery is intriguing, but the mystery isn’t solved yet. Watch now!
#AI #Science #Biology #Genetics
Subscribe for more fascinating science.
Comment what you liked most about the video!
👉 This channel was created in collaboration with https://www.youtube.com/@realidadeimpressionante
• The Inspired Business Technology de RomanSenykMusic
0:00 - AI Agents Discover the ARTE System
2:26 - ARTE’s Clues and CRISPR’s Origins
4:52 - How CRISPR Became a Gene-Editing Tool
9:59 - ARTE and AI’s Expanding Role
11:11 - AI Predicts and Designs Proteins
15:43 - Genetic Medicine and Human Enhancement
19:42 - The Ethics of AI-Accelerated Genetics
#AI #Science #Biology #Genetics
Subscribe for more fascinating science.
Comment what you liked most about the video!
👉 This channel was created in collaboration with https://www.youtube.com/@realidadeimpressionante
• The Inspired Business Technology de RomanSenykMusic
0:00 - AI Agents Discover the ARTE System
2:26 - ARTE’s Clues and CRISPR’s Origins
4:52 - How CRISPR Became a Gene-Editing Tool
9:59 - ARTE and AI’s Expanding Role
11:11 - AI Predicts and Designs Proteins
15:43 - Genetic Medicine and Human Enhancement
19:42 - The Ethics of AI-Accelerated Genetics
Category
🤖
TechTranscript
00:00For 21 hours, nearly a thousand artificial intelligence agents investigated more than
00:06200,000 enzymes that are present in nature, and they found something that was hidden in
00:11the DNA and that scientists still can't fully explain.
00:15The discovery has similarities to CRISPR, that technology that already allows us to
00:20modify human genes and treat diseases that were once considered incurable.
00:25But now artificial intelligence is going beyond analyzing the code of life.
00:30It's helping scientists design new biological structures, and this brings us closer to
00:35a possibility that humanity has been pursuing for a long time, you see?
00:40Living longer, slowing down the effects of aging, and who knows, maybe one day overcoming
00:45the biological limits of our own bodies.
00:48But are we really prepared for what happens when we start to rewrite our own lives?
00:54That's what we're going to understand from now on.
01:02For about 21 hours, to be more exact, right, 950, artificial intelligence agents worked
01:09in parallel, investigating more than 200 enzymes associated with genetic mechanisms found in
01:15nature.
01:15They searched for patterns, compared sequences, and explored possibilities that could reveal
01:21biological systems that are still not well understood by us.
01:24And in the middle of all this investigation, something unexpected came up.
01:29A set of genetic sequences with an organization that resembled CRISPR in some respects, that system
01:35that revolutionized our ability to edit DNA.
01:38The discovery was named RT, but there was a problem.
01:42No one knew exactly what that system was doing at that moment, and that's exactly where this
01:47story starts to get really interesting, you know?
01:49Because the result didn't come from a traditional experiment conducted step by step by a team
01:54of humans.
01:55It came from a large-scale investigation in which hundreds of artificial intelligence agents
02:00explored different paths until they identified a candidate that deserved a lot of scientific
02:06attention.
02:07For the first time, we are now beginning to see much more clearly what happens when an artificial
02:12intelligence stops just answering questions about biology.
02:16And now it starts to help formulate new hypotheses about the mechanisms of our own life.
02:21And what caught the researchers' attention wasn't a completely new enzyme, okay?
02:27Reverse transcriptase had already appeared in previous studies about a type of bacterial
02:31virus known as bacteriophage.
02:34The novelty was in what existed around it.
02:38Clouds agents examined the DNA near the gene for this enzyme and found a long sequence of
02:42repeats organized at regular intervals.
02:45Next to it, there was also a gene associated with another protein whose function was still
02:50unknown.
02:51Separately, these two elements might seem like just a few more sequences among billions of
02:56pieces of genetic information.
02:58But together, they formed a very common architecture.
03:01The agent began to count their repetitions, measure the intervals between them, compare the
03:05organization with already known systems, and look for possible descriptions in existing
03:10scientific literature written by humans.
03:12That's when the hypothesis arose that those sequences could be part of a biological mechanism
03:17that hadn't been characterized yet.
03:20And the first laboratory test brought an additional clue.
03:24The repetitive region of ART produced small RNA molecules.
03:28This caught attention because CRISPR also uses RNA molecules derived from repetitive sequences
03:33to guide part of its function.
03:42So far, researchers have not been able to prove that ART can cut, copy, or insert genetic material
03:47like other molecular tools, such as CRISPR.
03:50They also still don't know what its natural function is within bacterial phages.
03:54For now, we are dealing with a scientific hypothesis, which is very promising, but it's still just
04:01a hypothesis.
04:01And it still needs to go through further experiments.
04:04And there's another important detail to mention, okay?
04:07Although Cloud conducted the computational investigation and helped identify the candidates, the physical
04:12experiments are still being carried out by human scientists.
04:14Even so, the case caught the attention of Fang Zihang, a researcher at MIT and a pioneer in gene editing
04:21like CRISPR, who considered the association between those repetitive RNA sequences and reverse
04:26transcriptase is something that definitely deserves investigation.
04:29Because the story of CRISPR itself began in a similar way, with researchers trying to understand
04:34genetic sequences that were strange and found in microorganisms.
04:38For years, those repeated sequences seemed like just a curiosity of nature, until scientists discovered
04:44that they were part of a sophisticated bacterial defense system.
04:48And what happened next completely transformed our relationship with DNA.
04:52And this is where we need to understand how a discovery made in bacteria ended up giving humanity
04:58one of the most powerful tools ever developed to modify the very code of life.
05:09And, as mentioned, for several decades, human scientists found seemingly strange sequences
05:15in the DNA of bacteria.
05:16They were small stretches that, repeated and separated by genetic fragments, were different
05:21from each other.
05:22At first, no one knew exactly what all of that was for.
05:25But there was something curious about that organization.
05:28Some of the sequences found between the repeats were almost identical to fragments of the genetic
05:33material of viruses that attack the bacteria.
05:36It was as if those microorganisms were keeping pieces of their own enemies inside their own DNA.
05:42And that was exactly the missing piece.
05:45When certain viruses infected a bacterium, they introduced their genetic material into the cell
05:50and tried to use its structure to produce new viral particles.
05:53But some bacteria simply developed an extraordinary defense mechanism.
05:58They are able to incorporate small fragments of the invader's DNA into their own genome.
06:03And if the same virus tries to infect them again in the future, they use that information
06:08to recognize it.
06:09The system was named CRISPR, an acronym that describes exactly the organization of these
06:15repetitive sequences.
06:16But the most interesting part was how this defense works.
06:19The bacterium produces small RNA molecules from the information stored in the CRISPR system.
06:24These molecules act as guides, helping specialized proteins to locate matching sequences in the
06:29invading genetic material.
06:30In some of these systems, a protein called Case9 acts as a kind of molecular scissors.
06:37When the guide RNA finds a sequence that matches the virus's DNA, Case9 can make a cut in that region,
06:43only compromising the invader's ability to continue its infection.
06:47In other words, the bacterium had developed a mechanism capable of recognizing and cutting genetic material
06:52with impressive precision.
06:54And that's exactly when the researchers realized, something that would completely change the
06:58meaning of that discovery.
06:59If an RNA molecule could guide Cas9 to a specific DNA sequence, maybe it would be possible to
07:06modify this RNA to direct the protein to others.
07:10Others, but not just virus genes, but plant genes, animal genes, and even our own genes.
07:16In 2012, Jennifer Doudna, Emmanuel Charpentier, and their collaborators demonstrated how this
07:22mechanism could be adapted to make programmable cuts in DNA sequences.
07:26It was the beginning of a transformation in genetic engineering.
07:30Until then, modifying a specific gene required complex tools developed individually for each
07:35situation.
07:36But with CRISPR-Cas9, much of this difficulty started to be solved by altering the RNA sequence
07:42that guides the protein itself.
07:44Instead of building a completely new tool for each region of the genome, scientists could now
07:49use a similar molecular system to modify these instructions.
07:53DNA had become much more accessible to human intervention.
07:57In other words, easier for humans to modify.
08:00But cutting a genetic sequence is only part of that process.
08:04After the DNA is cut, the cell itself tries to repair the damage.
08:08Depending on the strategy used by the researchers, this repair can deactivate the gene or allow specific
08:14changes in its sequence.
08:15And that's exactly where some of the greatest possibilities and also the greatest challenges
08:19of genetic editing arise.
08:22Because DNA doesn't work like regular text, where you can just erase a word and write another
08:28one.
08:28Genes interact with each other, take part in complex processes, and can influence different
08:33characteristics of an organism.
08:35Besides that, the editing system needs to reach the correct cells, as well as recognize the
08:39desired target and avoid unwanted modifications in other regions of the genome.
08:44And from that discovery, modifying the virus's code became much easier.
08:48However, predicting all the consequences of these modifications still isn't.
08:52Even so, this breakthrough was so important that, in 2020, Doudna and Charpentier received
08:58the Nobel Prize in Chemistry for developing a genome editing method.
09:01And a few years later, this technology stopped being just a laboratory tool.
09:05In 2023, the regulatory authorities of the United Kingdom and also the United States approved
09:11one of the first therapies based on CRISPR.
09:13In this case, it was a treatment for sickle cell disease, a hereditary condition that affects
09:18the function of red blood cells.
09:20In this procedure, blood-producing stem cells are taken from the patient and modified in the
09:24laboratory.
09:25The goal is to reactivate the production of fetal hemoglobin, helping to reduce the complications
09:29caused by the disease.
09:31Then, the modified cells are returned to the body.
09:34It's not about rewriting a person's entire DNA.
09:36It's a targeted intervention, performed on specific cells with a defined therapeutic goal.
09:42But it represents something that, for a long time, seemed to belong only to science fiction,
09:47a genetic disease being treated by editing human cells to change how they function.
09:52And all of this started with researchers trying to understand those strange repetitive sequences
09:57found in bacteria.
09:59That's why the discovery investigated by the cloud artificial intelligence agents is attracting
10:03a lot of attention worldwide.
10:04ART also shows a common genetic organization, like the repetitive sequences and associated
10:11RNA molecules.
10:12But that still doesn't mean it works like CRISPR, okay?
10:15Nor that it can be turned into a similar tool.
10:17There's still no proof of anything yet.
10:19Its function still needs to be thoroughly studied and demonstrated in the lab.
10:23However, the connection between the two hypotheses lies in something else.
10:27Nature contains molecular mechanisms that, yes, it can take us decades to understand.
10:32Even after identifying the genes involved.
10:34And now artificial intelligence is already starting to take part in this search and speed
10:39all of this up.
10:40But it's always good to mention that there's a very important difference between the moment
10:44when scientists discovered CRISPR and what's happening today.
10:48AI is no longer being used just to look for mechanisms that nature has already developed.
10:53It's also now starting to help researchers design biological structures that may have never existed
10:59before, okay?
11:00And it's with this change that we enter a new stage in history.
11:04Let's understand this.
11:11Over the supposed billions of years, nature has developed an extraordinary amount of proteins.
11:17There are molecules responsible for practically everything that happens inside a living organism,
11:22transporting substances, transmitting signals, fighting infections, building cellular structures,
11:28and carrying out chemical reactions essential to life.
11:31And each of these proteins starts with a sequence of amino acids organized according to the instructions
11:35stored inside the DNA.
11:36But there is a fundamental detail.
11:39To understand what a protein does, it's not enough to just know its sequence.
11:43You need to find out the shape it takes in space.
11:46And this happens because a protein doesn't remain as a simple chain of amino acids.
11:50It folds into complex three-dimensional structures.
11:53And the way these structures are organized determines a large part of their functions.
11:57In the past, it was discovered that figuring out the structure of a single protein could require
12:02months or years of experimental work.
12:04It was exactly with this problem that artificial intelligence caused one of the biggest recent
12:09transformations in biology in 2020 AlphaFold, three-dimensional.
12:12Developed by Google's DeepMind, demonstrated an extraordinary ability to predict protein structures
12:18from their amino acid sequences.
12:20Two years later, DeepMind and the European Bioinformatics Institute made structural predictions available
12:26for more than 200 million proteins.
12:28And suddenly, researchers from different parts of the planet gained access to information
12:32that previously could have required decades of experimental investigation.
12:36But predicting the structure of a protein is different from fully understanding how it works.
12:41A protein structure doesn't automatically reveal all the molecular interactions,
12:45the movements, or the chemical reactions that that protein can perform.
12:49Even so, it offers something that is extremely valuable,
12:52a starting point for investigating biological mechanisms that were previously very difficult
12:57to study.
12:59And it was precisely from this capability of artificial intelligence that an even more
13:03ambitious question arose, if AI can learn how proteins are organized,
13:08can it also help create completely new proteins?
13:11The answer began to appear in different laboratories around the world.
13:15Researchers at the University of Washington, for example, developed an artificial intelligence
13:18system that is capable of generating protein structures based on specified design goals.
13:22One of these systems is known as RF diffusion,
13:26and it uses principles similar to those employed by generative image models.
13:30But instead of producing photographs or illustrations,
13:34it proposes three-dimensional molecular structures.
13:36Scientists can define certain desired characteristics,
13:40such as the geometry capable of interacting with another protein,
13:43and from there, use that model to generate candidates.
13:46Then, these structures need to be translated into amino acid sequences,
13:49are physically produced, and also subjected to experiments, of course.
13:53And it's an important change, because artificial intelligence is not just looking for solutions
13:57that already exist in nature, it's helping to imagine these new molecular possibilities.
14:02And in 2024, another breakthrough drew a lot of attention.
14:07Researchers introduced ESM-3, an artificial intelligence model trained to work with information
14:13about the sequences, structures, and also the functions of proteins.
14:16And during the experiment, the system was used to generate a new variant of fluorescent protein.
14:21These proteins are especially important, because they allow us to visualize biological processes.
14:26When used as markers, they can help scientists track cells and microscopic structures.
14:32The protein produced with the help of ESM-3 showed this fluorescence in experimental tests,
14:36even though its sequence differed significantly from those of previously known fluorescent proteins.
14:40So what?
14:42Scientists who developed this new protein estimated that,
14:45to reach a similar difference through natural evolutionary processes,
14:49it could take hundreds of millions of years,
14:51although this comparison is just their estimate and not a direct measure of evolutionary time.
14:57Alright?
14:58The result doesn't mean that artificial intelligence has created life just to make that clear,
15:04but it shows that computational models can propose functional molecules
15:07that don't simply match sequences already found in nature.
15:11That's it.
15:12Protein.
15:13And this is where the story starts to connect again to everything we've been seeing.
15:18The experiment with the clod agents represents an AI searching for molecular biological mechanisms.
15:23Still not well understood.
15:25Meanwhile, AlphaFold represents the ability to predict molecular structures,
15:29and the generative protein models represent a next step,
15:32proposing new structures for human scientists to investigate their function.
15:36Discovering, understanding, and starting to design is just the beginning.
15:40These are three very different abilities, but they're coming together in the labs, you see?
15:44And that brings us to a consequence that goes beyond protein research.
15:50If we can use artificial intelligence to study genes, design molecules, and improve tools capable of modifying cells,
15:59how far can this technology go when applied to the human body itself?
16:03That's a question that deserves a lot of attention, doesn't it?
16:06Because it's one thing to use genetic engineering to treat a disease.
16:10It's quite another to start asking if we could modify human traits that were never considered diseases.
16:16And it's exactly on this border between treatment and enhancement that the next chapter of our story begins.
16:28The idea of modifying human DNA was just a possibility limited to laboratories until recently.
16:35But advances in genetic engineering are already starting to turn this idea into real treatments, you see?
16:43And artificial intelligence can speed up this process even more.
16:45Today, there are already therapies capable of modifying human cells to fight certain diseases.
16:50Besides CRISPR, researchers use cell engineering techniques to develop treatments for some types of cancer,
16:56like CAR-T therapies, for example, in which the patient's own immune cells are modified to recognize and attack tumor
17:03cells.
17:04But there is an important difference between treating a disease and modifying a human trait.
17:09It's exactly at this boundary that genetic engineering now begins to raise much bigger questions.
17:15Researchers are already investigating strategies to correct mutations associated with hereditary diseases
17:21to modify immune system cells and also to stimulate tissue repair mechanisms.
17:26In regenerative medicine, for example, scientists study ways to control the molecular signals that guide the growth and recovery of
17:33cells.
17:34Some animals, like salamanders, are able to regenerate entire limbs.
17:39We humans, on the other hand, have recovery mechanisms that are much more limited.
17:44Understanding these differences can help in the development of new therapies,
17:48although we are still far from being able to regenerate an entire arm, for example.
17:53And that's where artificial intelligence starts to take part in the research,
17:57helping to identify genes, proteins, and combinations of cellular signals that improve.
18:02The possibilities don't end with the recovery of lost functions, you know?
18:06And what if in the future these tools were used to modify human abilities that are considered normal,
18:11increase resistance to certain diseases, alter muscle characteristics,
18:15or interfere with processes related to aging?
18:18All these are possibilities being investigated at different levels by different teams of human researchers
18:23using artificial intelligence, but of course, they still don't represent safe and proven human enhancements.
18:31Characteristics like intelligence, strength, and longevity depend on complex interactions between countless genes
18:36as well as environment and biological development.
18:39There isn't a simple genetic switch that can turn a person into a superior version of themselves,
18:44at least as far as we know, right?
18:46And modifying a gene to obtain a certain benefit can cause unexpected consequences in other functions of the body,
18:53and that's why genetic engineering applied to human enhancement remains much more limited than some predictions suggest, alright?
18:59Even so, the direction of this research reveals a very important change.
19:04For several decades, the main goal of genetic medicine was to understand and correct changes responsible for diseases.
19:10And now, with the help of artificial intelligence, which is evolving more and more,
19:14it is helping us discover and design biological mechanisms.
19:17And we started to discuss possibilities that go beyond restoring health.
19:22And when technology stops asking only how to heal the human body,
19:26and starts to explore how we could modify our own bodies,
19:30then an inevitable question arises, doesn't it?
19:32Who will decide how far we can go and who will benefit from it?
19:41Back in 2018, the Chinese scientist He Jiankui announced the birth of two girls
19:47whose embryos had been genetically modified using CRISPR.
19:51The stated goal was to alter a gene associated with the entry of HIV into human cells.
19:56The experiment, of course, sparked an international reaction.
19:59Besides the major ethical questions, there were also doubts about the safety of the modifications
20:03and about the possible consequences for the development of these children.
20:07For the first time, the world was faced with the birth of human beings
20:11whose DNA had been deliberately edited even before they were born.
20:15The case showed that a technology capable of treating diseases
20:19could also be used to make hereditary changes
20:21that could potentially be passed on to future generations.
20:24And now, with artificial intelligence accelerating the discovery of genetic mechanisms
20:29and the development of molecular tools, this discussion takes on a new dimension
20:33because the speed of innovation may begin to outpace our ability to assess these consequences.
20:39An artificial intelligence can identify thousands of possibilities for genetic modification,
20:44but determining which of them are safe, necessary, or acceptable
20:47remains a human responsibility.
20:50There's also the question of access,
20:52if the technology is capable of modifying biological traits,
20:55and this becomes a reality in the coming months, years, or perhaps decades,
21:00who will be able to use it?
21:02Will these treatments be available to all human beings
21:05or just resources restricted to a small portion of the population who can afford them?
21:09How can we prevent interventions meant to fight diseases
21:12from ending up being used to establish standards for human traits
21:16that are considered desirable, like eye color, hair type, hair color, and so on?
21:22These questions still don't have definitive answers, do they?
21:25I think I already have, but there's an important difference between what happened back in 2018
21:30and the scenario that's starting to emerge right now.
21:33At that time, the debate was mainly focused on the human ability to modify DNA.
21:38Today, we also need to consider the role of this artificial intelligence system,
21:42which can already help discover, design, and improve these tools.
21:47AI is not taking control of human evolution, at least that's what they say,
21:51but it is starting to expand the possibilities of intervening in the mechanisms of life.
21:56And maybe that's exactly the most important consequence of this discovery
22:00that started the whole story we're telling here.
22:02Because hundreds of cloud agents investigated genetic sequences
22:06and found a biological system whose function we still don't understand.
22:11For starters, we don't know if ART will have practical applications.
22:15Maybe we'll never become a genetic engineering tool like CRISPR, for example,
22:18but the experiment points to something bigger, you see?
22:20Now, for the first time in human history,
22:23we have machines capable of participating in this investigation
22:25on a scale that would be extremely difficult to achieve with human work alone.
22:28And as we learn to understand and modify these mechanisms,
22:32perhaps the most important question will no longer be
22:34what artificial intelligence will be able to discover about life?
22:39About what?
22:41About life, actually.
22:43And now it becomes about what we will decide to do with all these discoveries.
22:48If you enjoyed this episode, you know what to do, right?
22:50Check that you're subscribed, give the video a like,
22:51and help me by sharing this video if you can.
22:53On your screen, there are two new suggestions,
22:55one here and another here.
22:57Click on one of them and I'll see you in the next one, okay?
22:59Thanks, bye-bye, and see you soon.
23:00I'll be waiting for you.
23:00Bye-bye, and see you soon.