jfarlow

Yep - and because DNA's base pairing rules are so well-studied, so predictable, and information-carying, we can use DNA for its material properties in addition to or even separate from its genetic properties. In terms of software, Shawn Douglas built CADNano [1] - software to do precisely that. By using DNA as a material it can be useful in its own right - with all sorts of interesting 3D structures, 3D logic, and with atomic precision, built into the encoded base pairs. But these structures generally do not interact with DNA at a genetic level in an organism.

In terms of protein design at that atomic level, the computation traditionally has relied on knowing or guessing at the structure (atomic arrangement) of the protein. And without that, there's not much to do (that's where our work picks up). A lot of that kind of protein design computation work is being done with software like Rosetta [2].

[1] https://cadnano.org/

[2] https://www.rosettacommons.org/

We do this at Serotiny [1]. And others are working on this as well for different specific purposes. We use some very basic ML (the data is very sparse, wide, but any hit above a low baseline is very valuable).

The curious part, from our perspective is that biology has massive surface area - and the surface area is 3D. It not only scales between species/functions, but it also scales up and down, from atoms to organs. And the expertise/abstraction layers that work at one scale become complicated if you try too hard to account for all variation at a different scale. HIV's genome is a backwards, upside-down, mirrored, fugue of an engineering design that uses exotic molecules, exotic regulation, exotic proteins, and exotic physics. We're starting at a different place, just trying to write very simple scales.

In our case, we've chosen a single size scale to work with - proteins, but are wide enough to look across every species & discipline to understand those proteins as common tools. We compile all of our designs for a particular function that we're interested in, down to DNA, literally. And finding the niche where we do not have to deal with all of the DNA-regulation, or cellular regulation, or tissue synthesis, etc. allows us to expand and build in complexity at the protein level - while keeping other parts of biology constant. And that also allows us to interact and work with others who are working at different scales.

And there are others that build in complexity at other biological levels (gene regulation, pathway flux, etc.). Companies like Asimov [2] are involved in similar work at some of those abstraction layers. The open-source design language, SBOL is an attempt to standardize the DNA layer [3]. And this contributes to the challenge in that a lot of people/companies/labs have projects to build an abstraction layer that compiles down to DNA - but they might be talking past each other and be doing separate projects.

We've built an entire API of 'high-level' commands at an abstraction layer above DNA, where the output compiles down to literally, a JSON file specifying the DNA sequence to be manufactured by a 3rd party, as well as human-level citations to enable turning the new designs into intellectual property.

There is still a LOT of data missing, and there's a lot of empirical work to do - and you need to keep your compiling system constant enough that when you make changes at your abstraction layer you know that when you hit a roadblock you know it's because of a change you made, and not just a bug in the system.

[1] https://serotiny.bio

[2] https://www.asimov.io/

[3] http://sbolstandard.org/

Don't let it make you mad - this is science. And this kind of refining progress is precisely how science should work. This is a good thing. Start with a toehold, and move inward. Cycle back when you realize your toe wasn't in quite the right place. But you couldn't have even gotten there without that toehold. We need controlled model systems to do science, even if the models are imperfect. And that's okay - the (good) scientists know this.

If we had started with 'full' media from the beginning we'd still be describing and making media variants rather than getting any cancer research done. And we've done a hell of a lot of good cancer research since these media were defined. It's a little like the 'Standards' xkcd comic - if you make too many standards, the standards cease to be useful as a standard.

Further, the effect of the media in this case here is likely more profound when doing 'biology' rather than 'biochemistry' experiments. And until recently, the cells themselves were toeholds onto real human biology; HeLa cells are very unusual, but we used them because we could reproduce the work between labs because the cells proliferated indefinitely. Who cares about a little unnatural media when the genomes of the cells are completely messed up. (And again, the use of HeLa cells taught us a LOT even if they were very not-normal). Normal human cells just don't last outside of a body - so if one insisted we only use normal human cells there would be zero research, and even less reproducible research because no one could use the same stocks.

It's only and precisely because we do have a (admittedly flawed) standard that such a different and likely more natural media even can be objectively compared against.

Biology has a lot of variables. You can't hold everything constant the way you'd like. So you pick just those that you expect will make the biggest differences. And we should certainly go back and check ourselves every once in a while to ensure we don't get lazy and assume a particular protocol should remain the same in the face of changes in the field. But I don't think that's the problem here. We're just better scientists now - and we're learning that our early assumptions can now be improved on.

The article, especially in its final paragraphs is actually quite even handed. The title, however, is very much hyperbole.

Serotiny | Bay Area, CA | Lead Back-End Engineer (First SWE Hire) | Full Time

Serotiny is a synthetic biology design firm that invents custom proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics. See https://serotiny.bio/notes/proteins/

Industry: Synthetic Biology, Biotherapeutics, Bioinformatics, Genetics

Location: San Francisco ideally, limited remote

Stack: REST API written in Go, Javascript (Ember) front-end

Role: API engineer - refine & augment API architecture, develop data structures and algorithms, help invent data-standards for synBio

Perks: Small credentialed team, flexibility, community engagement

Benefits: Health, Dental, Vision

Compensation: $100-150k + Equity

https://serotiny.bio/careers/

Happy to chat. My name is Justin, and my email address is my name at serotiny.bio

I still remember when 'googol' was my 2nd grade answer to 'no, I like it thousands better, no millions better, no billions better, no googol better!'

And somehow that silly abstraction for "a big number number than yours" became a mega-behemoth of a corporation.

I am curious how much of this is a bioinformatics insight. If your dataset is looking under the lamp-post for coding sequences, then you'd be unlikely to pull out non-coding regions for analysis. However, if you're strictly empirical in your search for tumor-specific antigens, then such sequences shouldn't be too much of a surprise. Either way, it's a very curious fact of biology here.

On a separate note, the idea that unused viral sequences start showing up as neo-antigens is also interesting. Those old viruses in our genome likely have curious roles in other parts of our immune system such that a therapy targeting them specifically might have odd repercussions.

Regardless, having more cancer-specific (cancer-enriched?) identifiers is a good thing. And especially if used with synbio 'logic' like that found in bi-Specific CARs or SynNotch, these just add more circles to the Venn-diagram that is used to uniquely identify cancerous cells amongst our 'self'.

Machinery of Life is a fantastic introduction to an intuition for what proteins and the mechanics of their scale.

https://www.amazon.com/Machinery-Life-David-S-Goodsell/dp/03...

That kind of thing is generally thought of as machine-learning, or AI assisted "Drug Discovery", but there's not a great name for the field at large. Because there are so many different kinds of 'medicine' at this point, there are a number of companies that use similar tools for various definitions of medicine:

Small Molecules: http://www.twoxar.com/ & http://atomwise.com/

Antibodies: http://www.antiverse.io/

Multi-Domain proteins: https://serotiny.bio (my company)

It is our job at Serotiny. We take plain-language requests for proteins with novel functions (not just binds more tightly, or produces this enzymatic product over that one), and we use the catalog of existing natural protein components to build you a protein that has those desired capabilities.

Imagine any natural input that life can read (light, heat, glucose levels, hormone levels, force, etc.) and any natural output that life can produce (temp, colors, fluorescence, electrical impulse, etc.). For many of those options, we can design a novel protein that achieves a linkage between that I/O.

However, our approach to the problem is very much not like AlphaFolds - we don't try to scan the 20^600 space by changing individual amino acids, but rather we don't worry about folding or structure (too much) and instead play around with discreet functional modules that already exist in nature. Our approach is a bit more sociological than it is a simulation of physics/chemistry. But it works.

Optogenetics tools, CARs, SynNotches, BaseEditors are all curious examples, and there are many more coming online right now.

Serotiny | Bay Area, CA | Lead Back-End Engineer (First SWE Hire) | Full Time

Serotiny is a synthetic biology design firm that invents custom proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics. See https://serotiny.bio/notes/proteins/

Industry: Synthetic Biology, Biotherapeutics, Bioinformatics, Genetics

Location: San Francisco ideally, limited remote

Stack: REST API written in Go, Javascript (Ember) front-end

Role: API engineer - refine & augment API architecture, develop data structures and algorithms

Perks: Small credentialed team, flexibility, community engagement

Benefits: Health, Dental, Vision

Compensation: $100-150k + Equity

https://serotiny.bio/careers/

Happy to chat. My name is Justin, and my email address is my name at serotiny.bio

You're right! The paper itself is strongly agnostic to its analysis scale, while the Quanta article focuses entirely on DNA sequence.

> Our embedded computational network architecture can be interpreted either at the levels of cells (internal network of molecules such as genes, proteins, metabolites) or alternatively as group or module of cells effective performing input-output transformation of external signals relative to the module. Naturally, this does not exclude a nested architecture ranging from cells encapsulating molecular networks to interacting modules composed of cells

And in fact the study actually meshes quite nicely with the paper cited before showing the evolution of signaling networks in multi-cellular systems.

So, coming from the science side of things where we 'program' DNA in order to design genes not found in nature, it's clear evolution works at many scales simultaneously. And those scales each need to be accounted for in order to traverse the large search-space. And in regards to this particular problem, evolution did not grant the complexity of multi-cellular life by doing any kind of random walk across a sequence. Life had a few billion years to come up with 'useful components' as well as 'good' heuristics and tools that commonly arrive at useful components. And then it reused that code, and started running different genetic algorithms on the same underlying code, but at another abstraction level upwards.

It allowed those components that were useful to be duplicated, redone, and rearranged to form higher-order tools. Where the components remained relatively stable, but were remixed, redone, reordered, and otherwise reused - with only slight modification to lock them into their new tool.

We see this with the results of evolutionary modifications to DNA. One of the proteins responsible for multi-cellularity is itself a mashup of 4 or 5 proteins found in single-celled organisms, but where those smaller proteins are repeated, smashed together, and otherwise remixed to form a protein with a new function that was synergistic with the original components. And that synergy allowed for the higher-order complexity associated with multi-cellularity [1]. And once that took off and proteins started to get large, the proteins would themselves be split up into pathways - and then those pathways would get duplicated. Etc. And this is just one mechanism real life has used evolution to scan that underlying space more rapidly than random walks.

This is actually what we do at Serotiny [2] - we don't design novel proteins by sequence (in the 20^300 landscape) but rather by functional domain (in the 20000^5 landscape). And I think there are mechanisms to incorporate that kind of scaling layer into a good genetic algorithm that allow you to traverse short paths by going upwards into the different space, and then coming back down when you need granularity,

And so, I think when going back to the algorithms, if you want to make them pragmatic rather than theoretical, many scales should be addressed simultaneously.

> “DNA is extremely important, but it makes no sense if [it is] not in a cell, in an organism, in an ecosystem.” Other interactions are at play, and this application of algorithmic information cannot capture the extent of that complexity.

Yep.

[1] https://www.ncbi.nlm.nih.gov/pubmed/11466531

[2] https://serotiny.bio

Serotiny | Bay Area, CA | Lead Back-End Engineer (First Software Engineer Hire) | Full Time

We're a recently funded synthetic biology company that has built the beginnings a biologically-aware API for designing novel genetic machines.

At Serotiny we invent proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics. See https://serotiny.bio/notes/proteins/

We're looking for a first hire who wants to help architect our digital infrastructure, up to and including developing novel synthetic biology data structures and algorithms for working with them.

Our API is currently written in Go - the front-end is written with Ember. Ideally you have shipped code, want to build a company's digital infrastructure, and want to learn some biochemistry (prior knowledge not required).

https://serotiny.bio/careers/

- High level article [1]

- A "Roadmap" from the semiconductor/data side, looking at DNA from its first principles [2]

- Review combining the tech side with the bio side [3]

[1] http://www.sciencemag.org/news/2017/03/dna-could-store-all-w...

[2] https://www.src.org/program/grc/semisynbio/ [see PDF on right side of page]

[3] https://www.biorxiv.org/content/biorxiv/early/2017/03/07/114...

Yes. Many orders of magnitude. However, if you're just doing cold storage, the data-density of DNA is many orders of magnitude greater. With essentially sub-nanometer 3D atomic data-density you can pack a staggering amount of data in a very very small package.

And though we're not quite there yet in read/write speeds at the ends of the route, the old saying rings true about the fastest way to transfer large amounts of data between the coasts - in a station-wagon full of hard drives.

This is already being deployed in therapies today - https://en.wikipedia.org/wiki/Checkpoint_inhibitor

And it works well in conjunction with other therapies that guide or otherwise help the immune system find the cancer. In particular, a style of designer gene therapy just approved by the FDA, of which one is called a Chimeric Antigen Receptor - https://en.wikipedia.org/wiki/Chimeric_antigen_receptor

These are in patients now, and curing people today.

If, for instance, the chimeric receptor (a designed synthetic cell-receptor) is able to enable immune cells to better sense the cancer, and that is coupled with the checkpoint inhibitors to prevent the cancer from disabling those immune cells, one's own immune cells will then be able to directly kill the cancer.

The checkpoint inhibitors (associated with the Nobel here), take the brakes off of the immune system. This can of course be dangerous, but generally one of cancer's first tricks it learns is how to apply those brakes to hide from detection. So by tuning down the sensitivity to the immune system's 'brake pedal', one's own immune system becomes more capable in fighting cancer.

Serotiny | Bay Area, CA | Lead Back-End Engineer (First Software Engineer Hire) | Full Time

We're a recently funded synthetic biology company that has built the beginnings a biologically-aware API for designing novel genetic machines.

At Serotiny we invent proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics. See https://serotiny.bio/notes/proteins/

We're looking for a first hire who wants to help architect our digital infrastructure, up to and including developing novel synthetic biology data structures and algorithms for working with them.

Our API is currently written in Go - the front-end is written with Ember. Ideally you have shipped code, want to build a company's digital infrastructure, and want to learn some biochemistry (prior knowledge not required).

https://serotiny.bio/careers/

Those "physical attributes" are most grossly characterized with those same measures- size, shape, length, stiffness, etc. And if you have access to those simple attributes you get a lot farther than if you just think about a genome as a magical information store.

My email address is my first name at serotiny.bio

-Justin

Serotiny | Bay Area, CA | Lead Back-End Engineer (First Hire, full-time)

We're a recently funded synthetic biology company that has built the beginnings a biologically-aware API for designing novel genetic machines.

At Serotiny we invent proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics. See https://serotiny.bio/notes/proteins/

We're looking for a first hire who wants to help architect our digital infrastructure, up to and including developing novel synthetic biology data structures and algorithms for working with them.

Our API is currently written in Go - the front-end is written with Ember. Ideally you have shipped code, want to build a company's digital infrastructure, and want to learn some biochemistry (prior knowledge not required).

https://serotiny.bio/careers/

I'd say, arguably, yes. And not necessarily because those examples use numeracy directly - but rather because a biological numeracy would provide a foundation to an intuition that would prevent the beginner's mistakes in those examples.

I argue that that biological numeracy is a prerequisite (and a consequence) of having a properly scaled biological intuition. And those examples test that intuition.

This is one of the most important parts of biological intuition that is very hard to get access to. Even PhD graduates often stumble when asked even about the order-of-magnitude of some of these values.

It's not dissimilar to the general order-of-magnitude of speeds associated with various computer components (RAM, CPU, Cache, HD seek, network latency, etc.). But there are many more components in a biological system, and their interactions up and down the physical scales are sometimes not obvious.

A simple example of intuition: once you've sequence your DNA, how would you go about changing it? Which DNA? Is the 'sequence' you got back an average sequence across just those cells that were sampled? Is the DNA you would want to change both copies, in each cell, in every tissue?

Another example is running through how a PCR reaction works. The actual result of the process often trips people up the first time because it's not immediately clear that there is an infinite pool of each component in the system to permit self-reactions in later steps of the process. Once a component is used in the first part of the process, it's still there in later parts of the process.

Very fun.

I put in some scientific proteins, and it spits out their common fluorescence staining colors:

ZO1 (stain tight-junctions, often blue): https://picular.co/zo1 (see https://www.google.com/search?tbm=isch&source=hp&biw=1399&bi...)

hN1 (a protein usually stained red or green with GFP or mCherry): https://picular.co/hn1 (see https://www.thermofisher.com/antibody/product/HN1-Antibody-P...)

Yep, that's because CARs are one of the first medically relevant genetic engineering projects that is not just a "cut-and-paste change". It is the insertion of a new capability into the immune cell. That is of a very different kind than prior approaches. And it's why the movement forward here really will start to accelerate as there are more and more changes that come online that more resemble CARs - and are inventive of new functions rather than just a simple tweak.

Serotiny | Back-End Engineer (First Hire, full-time) | Bay Area, CA

We're a synthetic biology company that has built the beginnings a biologically-aware API for designing novel genetic machines.

At Serotiny we invent proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics.

The API is written in Go - the front-end is written with Ember. We're looking for someone who has shipped code, wants to build a company's digital infrastructure, and wants to learn some biochemistry.

https://serotiny.bio/careers/

Serotiny | Back-End Engineer (First Hire, full-time) | Bay Area, CA

We're a synthetic biology company that has built the beginnings a biologically-aware API for designing novel genetic machines.

At Serotiny we invent proteins to cure cancers and genetic diseases. We help design proteins involved in CAR-T therapies, CRISPR systems and other multi-domain protein therapeutics.

We're looking for someone who has shipped code, wants to build a company's digital infrastructure, and wants to learn some biochemistry. We currently work with Go and Ember.

https://serotiny.bio/careers/

It's been talked about a few times on HN already, but it's a very interesting protein. It suggests a genetic delivery mechanism that is already in the body - and thus doesn't invoke a lot of the immunogenic properties of using viruses for gene therapies.

If you want to play with the Arc protein check it out here: https://serotiny.bio/notes/proteins/arc/

This is a great project for a specific kind of problem - and one that's not the common target for silicon valley. It's hard to get good and useful data in places where connectivity is low. And the (coding and social) logistics of getting that data from ubiquitous non-smart-phones pose interesting challenges that ODK helps address.

It's also run by great people.

For those interested in seeing Glenda:

https://blog.golang.org/gopher