Both of these statements are still very much in flux:
"Affordable lab robots from companies like OpenTrons mean you can do batch experiments without hiring an army of people, and computational drug discovery from companies like Atomwise allows some experiments to be done completely in silico."
I work in this field and I strongly recommend being extremely cautious making claims like this. You can't just get rid of lab techs when you use OpenTrons (instead, they spend their time debugging the robots), and in silico discovery of drugs is still fairly unreliable.
I guess I don't understand how OpenTrons will make a big difference. The pipetting robots we have around my company are the "expensive" ones by Hamilton and other traditional makers. I think those machines cost $40k, which is definitely more expensive than OpenTrons, but they're hardly a limiting cost compared to generating samples and the disposables for whatever assay you're automating.
Also, we have a department whose job is to run high-throughput assays, and they spend a lot of time and money validating an assay before they run the real samples through.
In grad school we bought robots because we thought they would make us faster. With very few exceptions they didn't, I was still faster with a multipipettor than our robots. This was because:
1.) Getting the robot to behave means it has to see the same thing every time, that's not easy to accomplish, hence the validation people do.
2.) Even when the thing is the same as before, robots break down. Handling solvents with robots is tricky because different solvents have different viscosities. If the robot misses a well, how do you know?
If you have a team that knows the robot well, and it's set up well, then all these problems are manageable. But lets not kid ourselves that it's just trivially solved now because we have robots.
An analogy in server land, we now have AWS and GCP. Basically now it's trivial to automate things it would take weeks/months to do by hand in the past. However, they each have an army of sysadmins and engineers behind the scenes servicing all the "robots" and making sure everything runs smoothly.
Robots are absolutely useful, and major companies are building out massive robotics facilities to accelerate drug development. The point is that it's not a magic solution. It still takes a bunch of staff to run, and (much less appreciated) it takes a different mindset to be maximally useful.
For example, maybe instead of doing a dozen informative but slightly complex assays, you do 1000 less useful but less complex assays. The end result might be just as useful if not more, but you have to structure your experiment differently.
It's unlikely that technology will play out the same way in biotech as it has in pure tech. I'm heavily invested in the idea that it will be immensely useful, but given the different constraints and problem space I think the trajectory will be different.
That said, I do worry about all the hype and overpromising and whether that's eventually going to come back to haunt the entire field, like the (1st) AI Winter.
I do think that inexpensive liquid handling robots will make a big difference if they are used correctly. The expensive Hamilton robots are closer to ~$100k-$150k and require a ~$20k-$30k service contract per year. The opentrons robots are $4k total. That kind of price difference is a PhD-level scientist's salary. But you don't want to buy a cheap robot and soak up someone's time troubleshooting it. To effectively use something like an opentrons robot, you need a still-rare breed of scientist, someone who's both relatively competent at programming and also talented at the bench. This kind of person will become more and more common, but they are currently hard to find. Meanwhile, opentrons is working hard to lower the programming barrier-to-entry, but they're a ways off still.
From my experience buying/getting quotes for Hamilton/Tecan machines, they're more likely in the $150-250k range. At least for the ones much wider and capable than that OpenTrons appears to be.
My feeling is that OpenTrons and similar systems would be a cost-effective way to scale up to warehouse-scale high throughput biology with more control over the process automation, making it easier to (for example) upload all the generated data to a large cloud and do large-scale machine learning and analysis over it.
The model I like is basically the same as how Google scanned so many books so fast and cheaply- rather than buying the Hamilton equivalent, they built their own crappy scanners and trained the techs to deal with the crap. This was much more cost effective and scalable. but you need a ton of inhouse experts to deal with the crap hardware, and most large compannies don't want that- they just want a service that works.
Google could do that because the basic components were so cheap due to consumer imaging.
On the lab side, there’s no similar downward pressure for making more accurate automated pipettes cheaper - I fully believe it’s possible, but the market isn’t there. Also, with imaging books, you can get away with a lot since ML/cheap labor/etc can be thrown at the distortion correction problem. With lab tests, you can’t really accept that same amount of error (ok, you can iff you understand the distortion and can process more samples to compensate).
I do agree that most companies just want a service that works. The issue is that there’s no standardization. Depending on the assay you do, you will need wildly different setups and sensor systems. When you optimize whatever you are doing, often you have to re-tune your entire assay to bring it line with the new dynamic range of your output.
If that weren’t the case, companies like Emerald and Transcriptic would have succeeded years ago.
Selling things like DNA or cell lines have been the closest to a standard, scalable product. Otherwise, it’s just replacing armies of people at a CRO (Contract Research Org) with a custom project for automation and assaying (Zymergen’s model).
At least in the area I work in (ML on fluorescence microscopy images used to estimate cellular and molecular phenotypes), the problem is extremely similar to the book reader that Google made. Heck, when I worked at Google I even built (20% time, open source) a prototype that could easily have been improved into what I describe (https://github.com/google/cncmicroscope-cad). We have to do the same kinds of distortion corrections (well, actually ours are far harder to do properly) to make the images useful for ML.
I've always been sad that Emerald and Transcriptic haven't been insanely successful. I've tried to put my scopes into both systems (I know the founders), but never made any headway. I like OpenTrons better, but having spent enough of my time knee deep in tubing and stepper motors, I have more appreciation for well-engineered systems.
That's fair. I updated the text about OpenTrons. I didn't mean to suggest that all experiments could be automated. Today, as I understand it, it only really pays to automate workflows that you want to do many times.
Mostly it's hard to get from "here's a protein" to "here's a drug for that protein" for a variety of reasons.
1.) It's hard to know whether the candidate drug will bind. This is because while the physical principles are somewhat understood, they're way hard to compute and there are a lot of things that still need to be empirically determined. For example, if you have a protein structure, that's a single snapshot in a highly unusual environment for that protein (either a solid crystal or some form of NMR solution). It is often accurate, but a protein in a cell is always in thermal flux, getting knocked around by other proteins, and "breathes" so your structure may not reflect important biology. This also assumes you have a structure to go off of, but you don't always.
Models and computation are making progress in this particular area, but it's still nowhere near plug and play.
2.) Even if you have a drug that binds the protein of interest, it might be toxic, might bind to other things as well, might not last very long in the bloodstream, might not even get to the bloodstream in the first place if you take it orally. Tons of things can go wrong.
This area is still quite difficult for computation because the datasets aren't great and it's hard to get enough data to make progress.
3.) Assuming you have 1 and 2 covered, it's possible your protein is bad. Maybe it's not actually as important to the disease in practice and you don't actually alter patient outcomes.
For these reasons, in-silico drug discovery is still unreliable.
I'm working on a project that would try to address these shortcomings by taking a hybrid experimental-computational approach. Hopefully it works out :)
Need help? I'm a middle-aged computer engineer who went back to school for Biophysics. Just finished first year. Also have background in simulation for games.
The simulations can give you small steps, such as suggesting chemical alterations to your molecule that might have higher binding energy to a docking site that you're targeting, but that assumes that the docking site is stable, that there isn't nonspecific binding, and a whole host of other things, so the first thing you do with those possible changes is run them in a biological system, not a computer.
No one who knows what they're talking about would take the idea that you could do drug discovery entirely in a computer seriously at this point in time.
Problem with those editorial is that they are made by non-experts. Maybe if they had spend some time with OT or other liquid handling device they know what the deal is, and how wide the rift between reality and wishful thinking is.
I have mixed feelings about this. We are increasingly seeing "cool tech" in the biotech area, but it often doesn't look to me like it really supports better health per se.
Better health seems to come from "living right" and that would be better supported by companies that help people exercise, eat right, improve the hygiene of their homes, etc. Gyms, urban planning, worker's rights advocates and so forth seem to do a better job of actually serving a goal of better health.
But you don't really get to claim credit for dramatic health improvements via those pathways. You don't get to claim "Our gym helps reduce the incidence of cancer by preventing it!" In order to make impressive claims about curing cancer, you have to first wait for someone to have cancer, then fix it. Prevention isn't an exciting, drama-filled field. It's not where the big money and the impressive headlines live.
I have concerns that getting better at the biotech approach helps create a monster, but there's no stopping it because preaching about diet and lifestyle will never have the same sizzle as "curing cancer."
Life Sciences VCs that incubate and spin-out their own companies put serious muscle behind the startups, in two main areas: money and leadership. The article notes the first, going on to suggest large upfront financing is no longer necessary, but glosses over the second. In biotech it is still very much the status quo to have seasoned execs leading startups. Not because younger founders are incapable, but rather this is what pharma execs want to see, and have connections with. It's hard enough for a startup to get pilots and deals with pharmacos, exponentially so if the founders just took CS 270 with the CEO's grandkid.
I think that the phenotype of biotech founders is changing and will continue to change. Some of the early stage biotech VCs who built their businesses around funding ex-Genentech execs are beginning to fund younger founders (although they are still open to replacing them with experienced execs post Series A).
More $10B+ biotech companies have been built by younger (under 40) CEOs than experienced CEOs. The dogmatic preference for experienced CEOs in biotech is a relatively recent phenomenon (last 15-20 years). It is a function of 1) all the next-gen tech of the genomic bubble of the late 1990s flaming out (gene, cell and antisense therapy v1, genomics v1) and 2) the success of the asset-centric build-to-buy model in biotech VC.
If your model is to fund assets carved out from big pharma, develop them to human POC, then flip them back to big pharma, it makes sense to hire ex big pharma managers to run the company. If your model is to build a large, lasting startup, historical data suggest you are better off with a younger, more technical founder. In a world where pharma is not doing as much startup M&A, where capital is readily available from Series B to public markets, and where you can get drugs approved relatively quickly, more startups have the option of becoming independent companies and not just trying to sell to pharma
Yeah that's fair, from ppl I know at AI drug discovery companies it seems like selling to big pharma is tough. Potentially bc the people who assess the technical feasibility of your product may be put out of a job by it :/
Have you tried selling to startups? From what I've heard they are better customers for AI drug discovery services as there is less entrenched interest in manual med chem and they value lower cost / fast iteration more
AI drug discovery is not an ideal market -- crowded, and long horizons to validation checkpoints. Our value prop to pharma is in clinical trials utilities. Relative to drug discovery, better path from pilots to revenue deals, but still cumbersome. Nonetheless your advice to target startups is sound :)
Can you elaborate on "capital is readily available from Series B to public markets"?
I am not in this space myself, but I know a few founders who are and the perception is that they face meaningful headwinds from the late stage community due to their age and market preferences for asset-centric startups.
I should clarify that I'm referring specifically to companies developing FDA-regulated prescription medicines, I'm not as familiar with the device / diagnostics / "digital health" markets
There is more capital available in terms of number of dollars invested in biopharma startups. This is true across the board from Series A to IPO stage. 2018 was a record year for VC investment in biopharma. 2019 is down a bit but still shaping up to be the 2nd highest year on record [0]
There is a lot of Series A funding, I referred specifically to "Series B to IPO" because most Series A funding comes from 5-10 VCs who start companies in house. Later stage funding comes from a wider number of investors
The IPO market is also more open now than all but a handful of previous years.
There is also more venture investment in "platform", as opposed to asset-centric companies now than ever before [1]. Of the companies that went public since Jan 2018, ~25% of the programs they are working on are gene and cell therapy. Traditional small molecule programs represent under 50% of programs these companies are working on [2]. Historically essentially all FDA approved drugs are small molecules or biologics, so this shift to gene / cell therapy platforms is pretty significant
That said, a platform is only as valuable as the assets it generates. The value of a drug increases exponentially as it becomes "derisked" through clinical trials, and the value of preclinical or earlier programs is not super high. In many cases later stage VCs won't invest unless there is a fairly derisked asset, or unless there is some really compelling evidence validating the platform (Arvinas is a good example of platform tech that is well validated, they went public at preclinical stage and are worth ~$1B).
This is a function of the structure of risk in drug development and I think that it is rational for more advanced assets to be more valuable than less validated, but potentially more impactful platforms. I wrote an article on the relationships between risk and value in biotech that quantifies some of these ideas: https://www.baybridgebio.com/drug_valuation.html
There is also often a valuation disconnect between tech VCs, who may invest at earlier stages and at higher valuations, and biotech VCs who would ascribe lower valuations and / or require more derisking of clinical risk. The biotech VC market is very hot compared to historical activity, but it is not quite as hot as tech VC in general.
There is def some deep age bias in biotech VC, it sucks and I / my friends experience a lot of it. I think it's an irrational bias and it will get competed away.
Have the "Biotech Startup" companies changed much in biotech yet?
As the article describes there is sort of a merging of the concepts behind tech startups and biotech startups but some of the "biotech startups" I've seen seemed to be things like a fancy CRUD app for biotech and they just wanted "biotech" in their name to seem cutting edge. That makes it hard to know who "Biotech Startups" even are.
Other situations where we've heard about traditional tech startup culture applied to biotech where the leadership did not understand how incredibly iterative / random biology research can be and how that greatly influences the speed you can move at. Where you can take lots of data and come up with some results in tech as computers are predictable, biology tend to throw a lot of curve balls making all your data wrong when you least expect it.
Have "Biotech Startup's" show enough promise to get to this next stage in the first place?
It seems like the traditional Biotech industry is way more complex to break into than say putting a web-application product in someone's face.
The biotech startup world has changed a ton in the last 7 years. However, the biotech startup world is a completely different one than the tech startup world, and the tech press doesn't really talk much about biotech startups that aren't AI/ML related or funded by traditional tech VCs.
$17B was invested in biopharma startups in 2018 [0]. Less than 10% of that came from traditional "tech" VCs. Most of these companies are traditional drug development companies pursuing new ways to treat cancer, rare disease, autoimmune disease and neurodegenerative disease. If you're interested in learning more about these companies, i maintain a database of recently funded biotech startups with some stats on what type of companies get funded [6]
10 years ago, a biotech startup was basically an external R&D program for big pharma. If there was an asset or area of biology that was interesting but too risky for pharma, a VC would start a company to develop that asset until the big risks were taken out, and then big pharma would buy the company. The VC would hire big pharma managers and scientists to run the program, which makes sense as it is basically a carve out of a big pharma program
Today, startups are much more ambitious. They are developing powerful platforms that can support real companies, not just flipping assets to pharma. They are increasingly exploring new types of drugs: gene therapy, cell therapy, microbiome therapy, synthetic biology platforms. Pharma has no expertise in these areas, so some startups could build very valuable independent companies here. Many startups are now commercializing products on their own, which was unheard of 10 years ago.
Some of these companies are working on amazing science. There is so much exciting stuff happening in biology and bioengineering. The tech press focuses more on stuff like AI/ML, which is definitely interesting, but IMO that stuff is probably like the fourth or fifth coolest thing happening in biotech right now. When you read about AI/ML in traditional biotech press, most of the articles are about why it won't work. People are much more excited about other stuff
If you are interested in learning more about the biotech startup world, you can check out Endpoints News [1], FierceBiotech [2], and STAT [3]. These are some of the more popular startup news sources in biotech
I wrote a few articles about the "old" biotech startup model [4] and how that model is changing and will change in the future [5] if you're interested in more on what's changing now in biotech
As a pre finance guy, I feel like the funding landscape change has more to do with the massive liquidity post QE just sloshing around. I feel like a lot of the financial trends can all be traced back to this single factor. It's the reason why we've had such a massive bull run in stocks. Come next financial crises that could freeze up, but then rates are going to go back down to zero again and who knows what might happen. So Paul Graham could be right again.
Liquidity is definitely amping things up in biotech, but industry trends and scientific advances are playing equal if not larger roles.
The NASDAQ biotech index starting outpacing the more tech-focused NASDAQ and the S&P around 2012. Biotech's outperformance peaked in 2015 and has been growing roughly at the rate of the NASDAQ since. The growth around 2012 was driven by better than expected sales from a number of big biotech companies like REGN, CELG, GILD, AMGN, BIIB that drove their stocks way up.
Then from 2012-2015 biotech benefited from a ton of pharma M&A. 2012 was the "patent cliff" where a ton of blockbuster drugs went off patent. Some very big companies lost 30%+ of their sales. 2012 was the first year ever that the pharma industry's sales shrank. At the same time, R&D productivity in big pharma was declining and they were laying off tons of R&D staff, so they had no new drugs to replace the off-patent ones. So they started buying tons of startups, and paying high prices
M&A activity dropped after 2015, and the sector's growth has been more limited since.
There also seem to be fundamental advances in science that are decreasing the cost and risk of drug development. Some studies suggest that the probability of approval is actually increasing in recent years. Some of this is because FDA is becoming more supportive of innovation, some is because of advances in precision medicine, biology, biochemistry, etc
Interestingly very little of the current biotech boom has to do with AI/ML / novel "tech". There are certainly advances in computation, biostatistics etc that have enabled companies to develop drugs more efficiently, but these have mostly come from within industry or academia, rather than from tech startups. There are some really interesting tech startups working in bio that could drive a new wave of innovation, but that would be on top of the innovation coming from biologists, biochemists, biostatisticians etc
Another big change that plays into this trend is the switch in markets. Traditionally biotech was B2B focused (eg drugs, ingredients, biofuel commodity plays) but as the science platforms are being developed as articulated in this article we are seeing the beginnings of B2C biotech companies, eg Impossible Foods. I think this shift in the market is also going to drive very different fundraising strategies.
For people interested in Consumer Biotech, here are more orgs to look into: 23andme, Ancestry, FitBit, Garmin, Apple Watch/Health, Verily, Oxford Nanopore, czbiohub.
Given the huge overhang of real-world testing which vests in public health, I still beleive the best model for biotech in medicine and agribusiness is NOT the private sector, but I suspect I'm an outlier, if not actually a sole believer here.
Biotech is overly regulated today (for good reasons). Because of that the entrance barrier is very high. It costs a lot more money today to start and run a biotech company.
To run a clinical trial for FDA submission process will cost you from several $ million to $ 100+million. To bring a drug to the market is ~$1B.
Only big guys with deep pockets can afford the expenses. If anything biotech is getting consolidated and monopolized (similar to tech world where FANG own everything) just like the rest of the industries.
My prognostication for next 20 years - big pharma and a few other biotech giants will own everything promising in the biotech world.
About developing new pharmaceuticals: apparently something that works well with mice might or might not eventually work with humans, and it's a slow and costly process to notice this only at the end of phase III clinical trials.
If someone comes up with a way to do hyper-accurate and fast in silico testing of human physiology, the duration of the clinical research would be dramatically shortened from the 10-15 years it takes nowadays.
If this ever happens, the valuation of biotech companies creating pharmaceuticals will explode onto absurd levels (like internet companies did during the "IT bubble").
There is zero chance of this occurring in our lifetime. We don’t understand even a tiny fraction of human biology, and every new drug - by definition - is exploring a new part of chemical space that we’ve never sampled before.
This is the fundamental hard, hard thing about biology and medicine that silicon valley thought-leaders don’t understand. There is nothing about drug discovery that is deterministic.
Unfortunately this post does not give me much confidence that YC will replicate its success in biotech. To illustrate why I’m just going to focus on the first example they give, Shasqi, because coincidentally this is the type of idea I and my colleagues would regularly propose to each other and play “find the flaw.”
In this case, this falls into the bin of, “are you absolutely sure this is the reason why cancer therapies fail?” Localized drug delivery sounds great in paper, but to my knowledge and what I’ve learned talking to people who know more than me, its not the most important issue - its actually hard to even be absolutely sure that the effects of chemical chemotherapy are fully explained by tumor-local effects (there’s a huge immune system component to it), so its efficacy compared to real chemotherapy is hard to predict.
Their publication also reads like one of tens of thousands of similar papers and approaches that often get published in the hundreds or thousands of journals. The paper is not bad at all, its a good proof of concept of the _chemistry_ and the fact that it does at least something inside a living organism (as opposed to chemical methods and derivatives that are not even biocompatible), but it does not convince me a bit that this is actually a worthwhile method to think about.
While there are numerous issues that lead to it, the most important is the fact that this (and virtually every other) paper that tries to show efficacy in cancer therapy uses xenograft mouse models to show it; this is where you take mice which have most of their immune system knocked out and inject HUMAN cancer cells (that have been cultured in petridishes for decades) to make a tumor and then try out various drugs to see which cures the tumor. Why do we need to use immune-compromised mice? Because otherwise the mouse’s own immune system will recognize any cell that did’t originate from its own body as foreign and quickly kill it. So now, we have a system where this drug-modification is compared to its baseline (which as I mentioned above, is probably going to need a viable immune system to function as it does in a real patient) in a system that inherently stifles it. So even if it shows to be better in a wildly artificial model, it says nothing about its efficacy in a human. I’ll go out on a limb and say that there’s actually enough evidence in their papers that this will never be better than regular chemo in a human (will be glad to go into detail if anyone’s interested) but even otherwise there’s not enough info available to say that this will work in humans.
While that might have been more detailed than some might like, the point is that I’d personally not even consider myself an “expert” in cancer therapy; I merely did a PhD in a lab that used to do some stuff related to it. But I can already see so many red flags in this single effort (I’ve seen similar reflags, wherever I knew some stuff, in other YC funded companies too). So, at least for therapeutic areas, I’m still not convinced that optimistic outlooks towards such projects is a good thing.
Of course, just saying you shouldn’t be optimistic with biotech projects is not enough, and I’ve struggled till now to put it in reasonable words why biotech (especially therapeutic) is different from technology companies and projects. But this article solidified it in my head and I boil it down to two points:
1. In tech, when you show a proof of concept for a product, you have essentially created the final product in the final environment as its going to be used by at least some users. Almost always, the main things you need to figure out are how you can “scale” it, and how you can make it appeal to a wider audience; but the premise of the product is proven with the prototype. With biology however, the premise is most definitely not proven with anything short of a phase II clinical trial. Almost all animal models (or anything in a lab dish) are significantly different from humans in ways that are fundamentally important for the mechanisms of action for these drugs. The only way to be absolutely sure that a drug has a really good chance of working in a human being is to inject it into a human with that disease and see if they die. So no matter how fancy the xenograft model or what caliber the nature publication might be, its still at best comparable to a deck of powerpoint slides in the tech VC pitch analogy and not anything comparable to a real prototype.
2. In tech, its a reasonable assumption that you can build more or less anything as functionality in an app/website, as long as it makes reasonable physical sense, because many engineers can have a fully comprehensive mental model of what is and is not physically and computationally possible inside computers. In biology its not at all a good assumption that if it makes sense in your head, then it will work in a human being, because no single person (not even a mythical House MD) has a full mental model of what we currently understand about all our biological systems. Our systems are too complicated and intricate (and without an organized creator) to keep in head by anyone, so its really hard to predict what will fail and cause a problem with our hypothesis. To some extent, every biotech project we do, even as an expert, is akin to someone who has barely used a smartphone coming up with an app idea - of course once in a while its going to make sense and be brilliant, but most of the time they have no idea how technology works in the backend and hence cannot say what is and is not possible.
Given these fundamental differences between biotech and tech, I’m going to remain skeptical until someone tells how they have figured out a solution to them.
Thanks for reading the Shasqi paper and for the detailed comment!
It seems if I can paraphrase your argument that it boils down to: "Just because your drug works in a mouse doesn't mean it will work in a human. So even if it works in a mouse there is still a lot of technical and scientific risk left."
I certainly agree with that. But I don't think it's incompatible with what I wrote in the blog post. The way the economics of the pharmaceutical industry work, it is still worth taking drugs into human trials even if the chances of them working are small.
Definitely appreciate more people funding projects like this. Heck, hopefully I'll be applying with an idea like this in the future too!
My worry/question is, what steps are you taking to vet the eventual scientific feasibility of these ideas? My main thesis is that biotech ideas and projects are fundamentally different from tech in the sense that any prototype (less than a clinical trial) is not a real prototype so some method needs to exist to account for that in the evaluation.
I think you answered this very well yourself: "In biology its not at all a good assumption that if it makes sense in your head, then it will work in a human being, because no single person (not even a mythical House MD) has a full mental model of what we currently understand about all our biological systems."
We do our best to vet the eventual scientific feasibility of these ideas, but it's along the lines of "it makes sense in theory", not "it will definitely work".
Because, as you said, our understanding of biology is so incomplete, I think that is the best that is possible right now. But I think that's ok. We (as a society) should be pursuing all the ideas that make sense. Some will work out of the box, some will work with modifications, and some just won't work at all. But this is the only way I know to make progress.
These are YC style biotech startups which I don't think have played out yet. The traditional approach raises a few million to get to phase 1 and then usually gets bought if the data looks promising. Many of these have pharma partnerships and/or investment. Sometimes they IPO earlier to get some outside capital. What's good about that is that the public can get in on it sooner. The bio side is very much driven by academia and spin offs.
Depends on the tech. Aerospace has gotten significantly cheaper due to accurate simulation tools. You can design a supersonic airplane or rocket engine in software and expect the first prototype to at least sort-of work.
The improvement of tools and slowing of silicon process improvement means that a startup can fab a competitive special-purpose CPU. Hennessy and Patterson's Turing lecture has an overview [https://iscaconf.org/isca2018/turing_lecture.html]
Robotics has gotten easier and cheaper thanks to good off-the-shelf sensors and actuators.
The bio in biotech makes all the difference. Dismissing the unpredictable nature of bio R&D by saying we can just outsource to a CRO is a massive oversimplification. Time and effort is not linear to results in Biological research.
I completely agree that funding will change dramatically in the next few years, and that the phenotype of founders will change. YC is doing a lot of awesome things to lead this change.
Historically, most $10B+ biotherapeutics companies were started by young technical founders, but most VCs today start companies in house, then hire older, non technical execs from big pharma to run the companies. Most of these companies are built to sell, and will never be lasting companies. I wrote several articles on this [0], [1], [2], [3]
This model is poorly suited for companies based on innovative platform tech. Technical founders (biologists, bioengineers) are the right people for these companies. In this way, biotech is very much like the tech startup world pre-YC
However, I don't necessarily agree that the world is changing because starting a biotech company is getting cheaper. And I don't think the traditional tech funding model is right for biotech.
The biggest risk in biotech is that stuff that works in mice doesn't work in people. You need to filter out false positives early on to avoid spending $200M getting a drug that wont likely work to Phase 2. It is better to spend $2M properly vetting the science than it is to go cheap and fast and cut corners. The reason biotech VCs have done so well the last 10 years is that they are really good at finding bad science quickly. The mentality of "spend as little money as possible to quickly generate positive data to raise more money" is the mindset that leads to the reproducibility crisis in academia and, in extreme cases, to companies like theranos
Starting a biotech company is cheaper, but it has been cheaper for a while. The big change now is that biotech companies can get bigger because scientific innovations are letting us develop better medicines faster with higher probability of success. 10 years ago, it was unlikely for a biotherapeutics startup to exit for over $500M. Now we see dozens of $1b+ exits a year
Companies are getting bigger, and they are getting big faster. Kite, Juno and Avexis (gene and cell therapy leaders) reached $8-10B valuations faster than SpaceX, Twitter, Pinterest and Dropbox.
It is now possible to power law invest in biotech, which was impossible 10 years ago when you had to get venture returns for a fund in a world where the best exit you could hope for was $500M. In that world, VCs have to get consistent singles and doubles: low risk, low valuations. Now investors can take more risk (including management risk) and invest at higher valuations
But they need to take smart risks. Product cycles in biotech are still much more expensive and long than tech. You need to diligence the science and be thoughtful.
We treat the rest of the living world like things that we can use and dispose of as we see fit. We adopt an "I-It" stance to the living world around us only to discover that we are lonely and fundamentally dissatisfied, unhappy, unhealthy, plagued by disease, and rapidly destroying ourselves. It doesn't have to be this way. We can recover our "I-Thou" relationship with Nature.
You can talk to life.
It's not difficult. You don't need any technology at all to do it. You don't need any investment money.
You can literally communicate with other living things (and your own tissues†), exchange information, make arrangements, etc. At that point it's really easy to e.g. grow all your food and medicine in a garden. (Economically, of course, this would be radically disruptive at scale.) For example, they have never made a big deal about it but that's how Findhorn Garden started: they were (and still are) communicating with Nature Spirits https://en.wikipedia.org/wiki/Findhorn_Ecovillage
But the actual "protocols" or "algorithms" to learn to communicate with other living beings are very simple and require no technology.
So, can I start a startup? Of course not, there's no money in it. In fact, if more people would do what I do much of the economy would evaporate.
- - - -
FWIW, I'm in San Francisco. If anyone reading this wants to meet up and learn about bio-telepathy and accelerated healing and stuff let me know.
†The other day I burned my thumb taking a pan out of the oven. When I stopped hopping around I communicated with the tissues in my thumb to "shape" or "guide" the healing response, with the result that the pain went away and didn't come back and the burn healed rapidly without leaving a scar. This is not a surprising or unusual event in my life. (Burning myself was unusual, I'm usually pretty careful.) Accelerated healing without pain is trivial if you know how.
"Affordable lab robots from companies like OpenTrons mean you can do batch experiments without hiring an army of people, and computational drug discovery from companies like Atomwise allows some experiments to be done completely in silico."
I work in this field and I strongly recommend being extremely cautious making claims like this. You can't just get rid of lab techs when you use OpenTrons (instead, they spend their time debugging the robots), and in silico discovery of drugs is still fairly unreliable.