Rising Tides of Innovation

This is a post based on a curiosity drive (prompted by a book I was reading). Took a few hours of reading and some synthesis. I do not claim to be an expert in most of what’s being said below :)

As the saying goes, rising tide lifts all boats. AI is one of those rising “tides” that’s expected to “lift” our lives. I was curious to find out about the impact and interconnectedness of such major “tides” on other contemporary tides - synthetic biology and genetic engineering, Fusion tech (the holy grail of limitless & clean energy). I would have liked to cover space exploration and quantum computing. But that's for a different post.

So, does the rising tide of one impact the other significantly? And do we see an exponential increase in the pace and progress in each of them aided by the other.

This led me to a rabbit hole of reading. As expected this threw some interesting developments that I wasn’t aware of. Read below if you are curious as well.

Note – I had to split the post into 2 parts after seeing the content. This focus only on Synthetic Biology & AI

·?????Protein structure and folding

Knowing how proteins are structured, how they react with other molecules, is very important. It helps to understand biological functions and how some diseases that are triggered through malfunctioning proteins (such as some cancers, Alzheimer, Parkinson’s) can be treated by drug molecules. So, a pretty important function.

Before DeepMind’s AlphaFold, predicting the 3D structure of proteins was an extremely difficult and time consuming task. Experimental methods like X-ray crystallography and others apparently took years of work and millions in specialized equipment. Other computational approaches had limited success, with accuracy often below 60% for complex proteins.

At CASP14 in 2020, AlphaFold 2 demonstrated a median accuracy of around 90. With the increase in performance and accuracy, aplhafold 3 has enabled the prediction of over 200 million protein structures, including nearly all cataloged proteins known to science.

It doesn't stop there. “Alphafold3 is expected tp unlock new understanding and significantly improves accuracy in multiple key biomolecule classes, including ligands (small molecules), proteins, nucleic acids (DNA and RNA), and those containing post-translational modifications (PTMs). These different structure types and complexes are essential for understanding the biological mechanisms within the cell… The model’s expanded capabilities and performance can help accelerate biomedical breakthroughs and realize the next era of 'digital biology’”

It’s not just DeepMind, there are other models. So much so that, Nvidia’s BioNeMo now has?six new optimized, open-source models. There are more than 160 drug discovery programs based on models like this and 15 of that apparently are in clinical development.

·????? Open source gene editor

Another application I was surprised to see was Language models being applied to design functional proteins.

OpenCRISPR-1 is an AI-designed open source gene editor released by a company called Profluent. This company apparently used a proprietory LLM that was trained over26 terabase (each terabase is one trillion base pairs of DNA) of assembled microbial genomes and metagenomes.

Profluent Successfully Edits Human Genome with OpenCRISPR-1, the World’s First AI-Created and Open-Source Gene Editor. “With this launch, Profluent demonstrates the first successful precision editing of the human genome with customizable gene editors designed from scratch with AI…Our success points to a future where AI precisely designs what is needed to create a range of bespoke cures for disease."

It's not just Proficient doing this work, Salesforce AI has Progen. “We aim to engineer novel proteins, whether undiscovered or nonexistent in nature, by tailoring specific properties which could aid in curing disease and cleaning our planet.”

https://singularityhub.com/2024/04/25/this-ai-just-designed-a-more-precise-crispr-gene-editor-for-human-cells-from-scratch/

https://www.biopharmatrend.com/post/642-from-gene-editing-to-pathway-design-how-ai-is-transforming-synthetic-biology/

·????? Synthetic Biological Intelligence

Let’s look at the other way around – has synthetic biology contribute to development of AI? Is anyone looking to biologically inspired AI systems?

I learnt about Hebbian theory - in neuroscience that describes how neurons adapt during learning and memory formation… if two neurons frequently activate at the same time, the connection between them strengthens. They state it as “neurons that fire together, wire together”. It explains the synaptic plasticity

This makes future communication between these neurons more likely and efficient.

This has been broadly compared with weights in an ANN, especially backpropagation. But this idea of changing weights is only during the training. Biological systems – our brain, continuously do this and that explains endless learning and ?

This is a specific example of how biological experiments can inform and validate techniques for AI models, making them more robust and capable of handling complex and continuous learning activities.

But how do you learn this – There’s a project called DishBrain that among other thigs can help observe and learn from neuroscience. DishBrain used neurons from human and mouse brain cells seeded in microelectrode arrays. Apparently the neurons once they are seeded establish connections with each other and form a network. It reacts to electrical stimuli through the electrodes and record the neuronal responses. This becomes an observable environment to learn from how real neural network works and inspire Artificial neural network techniques.

I don’t know how real this will be beyond a lab experiment. But this is utterly fascinating nevertheless.

Going forward, Scientists are planning to build hybrid networks where such networks are integrated with real-world sensors and output devices. This will be a true hybrid neural network.

https://www.frontiersin.org/journals/science/articles/10.3389/fsci.2023.1017235/full

?

·????? Organoid Intelligence

Projects like Dishbrain while it sounds complex, there are other crazy ideas being experimented. There is a thing called Organoid Intelligence (Organic Intelligence? that sounds like you and me!) that deals with organic and biological neurons in creating bio-computational systems or biocomputers.

“organoids are a type of lab-grown cell-culture. Even though brain organoids aren’t ‘mini brains’, they share key aspects of brain function and structure such as neurons and other brain cells that are essential for cognitive functions like learning and memory. Also, whereas most cell cultures (e.g. dishbrain project) are flat, organoids have a three-dimensional structure.”

https://scitechdaily.com/organoid-intelligence-revolutionary-biocomputers-powered-by-human-brain-cells/

·????? Generative Biology

“application of artificial intelligence (AI) algorithms and machine learning tools to biological applications – "generative biology" is changing how we interface with biology and the outcomes it can provide.”

A cool example I found is something called Solein. It is a protein developed by a company called Solar Foods by using elements of synthetic biology. It feeds CO2, water and others to a microbe in a bio-reactor resulting in this protein mass. The amazing thing is they are in production and is used to make snack bars available in Singapore. https://www.tatlerasia.com/dining/food/fazer-launches-first-chocolate-bar-with-solein-protein-powder

Similarly, another fat alternative called OleoPro is made using a platform that uses artificial intelligence to rapidly generate unique protein insights and identify naturally found novel ingredients.

I think this will be an interesting area to watch out for given the focus on sustainable development models. https://www.weforum.org/agenda/2024/05/overcome-barriers-to-generative-biology/

·????? Training set for AI models

AI Alphafold now claims to be able to master all structures in the protein data bank (PDB). Synthetic biology allows for the design and creation of novel proteins and enzymes, which are structurally characterized and added to the PDB.

Researchers used AI trained on data from high-throughput screening of synthetic microbial libraries to identify a new antibiotic compound called halicin, which was effective against many drug-resistant strains.

In the next post, I will talk about the interconnectedness of AI and Nuclear Fusion tech.