How to Start a Biotech Company on a Budget
blog.ycombinator.com
blog.ycombinator.com
Intense domain knowledge + knowledge of programming + affordable automation is really a fantastic combination to have with biotech. When the software isn't advanced enough, you can do tricks with hardware. When the hardware isn't advanced enough, you can do tricks with biology. When the biology isn't advanced enough, you can do tricks with software.
It really does depend on connections, though. All of my early customers are folks I've known in the community for years, so I can get paid while working out all the kinks in the system.
Right now, Opentrons rules with their open source robots, so liquid handling robots are pretty well handled. Xarm makes good arms for interfacing with Opentrons, so then it comes to everything else necessary to get to full automation.
1. Automated plate hotels I need plate hotels to storage samples that interfaces with my robot arm and robot systems. This I'm basically using Theo Sanderson's designs for this https://www.thingiverse.com/thing:2668120 since buying them commercially is so darn expensive. I really need this for my system, so I am building em.
2. Plate sealer / unsealers My entire full automation pipeline is built around NOT needing a plate sealer or un sealer (something you can't do with the vast majority of screening systems). Automated versions of just the sealer itself cost more than 2 liquid handling robots and a robotic arm combined (and all it does is seal plates!) I think there is an opportunity for building simple but effective automated plate sealers and unsealers.
3. A train system I think Concentric (Ginkgo's COVID company) uses something similar, but basically a system that can use small trains to bring plates and reagents around a lab to different robotic stations. I'll probably just do that with lines of robotic arms, but I definitely spent quite a bit of time thinking of how to do this. (in my factorio megabase I really learned how to route trains, and I think it's exactly the same problem I face with my robotic work stations). I'm not sure how to scale the selling of that though.
I did that in my garage but after having to clean up a big contamination and debug experiments due to temperature fluctations, I realized there was a reason people use real labs.
Garages are terrible for biotech because of contamination problems. I have a bedroom that I've sealed quite well with 2 box fans constantly running filtering the air (tape a 20x20 filter to a box fan and it works great). When I work with media or anything I also use a flame. I still have some problems with contamination, but it is manageable.
There are also biological ways to get around contamination problems. So long as your media is clean, if you use a strain like Vibrio natriegens, it just outgrows literally everything, so it's not that big of a worry.
I have a few ruuvitags + a raspberry pi to monitor the temperature, and I live in Coastal California, so we don't have that much fluctuation. It hasn't mattered for my experiments, anyway, which are mainly just DNA cloning.
I am interested in automating the lab I am running now. We have an opentrons that we bought couple years ago that I never got around to using. But now that we are regularly getting a decent amount of samples in a week, I really want to get it going. Are you starting a lab automation service?
I agree with your point about plate sealer/unsealer. It is a huge pain point in automation. However, I am not sure if you can convince regulatory bodies to approve it. Are you doing some kind of mineral oil top layer to prevent evaporation? We've actually tried that before and it kinda worked, but the problem is that the mineral layer messes with the SPRI bead purification step. I am sure it can be solved with more testing, but it just wasn't worth the inconsistency at the time.
> Are you doing some kind of mineral oil top layer to prevent evaporation
The opposite: the process is pretty much based off of dessication (plus a little bit of tube opening by the bots for important chemicals). For thermocycling stuff, I'm planning on just using Opentrons module with the reusable seals. I really wanted to find a way to use reusable seals everywhere, but I just couldn't think of a way to do it well robotically. Thankfully, I don't need to convince the regulatory bodies of anything, since I'm just doing DNA cloning.
> We have an opentrons that we bought couple years ago If you need to do multichannel stuff, do yourself a favor and go buy Gen2 multichannels and toss your old Gen1. The Gen1 multichannels are hilariously unreliable, but the Gen2 are pretty solid. Love those bots overall though - I got two of em so I can work in parallel!
Are you open to discussing some kind of consulting arrangement between you and my company? If so, how do I reach out?
Desiccation is just for the storage steps. Otherwise I just plan on passing liquids around, compensating for any evaporation.
> Are you open to discussing some kind of consulting arrangement between you and my company?
Sure! Contact me at koeng101 [@] gmail.com and we can talk
EDIT> I mean involved w.r.t plate sealers/unsealers, not more generally.
We're recently opened up a position for a Staff Software Engineer (https://apply.workable.com/billiontoone/j/2F0405817C/).
If you're interested, please reach out to me directly at david+hn@billiontoone.com
My wife and I have a setup up in our house (it's all GRAS and Biosafety Level 1 stuff) and have mostly bought off ebay and govdeals. She's a neuroscientist and I'm a physics/engineering guy. My job is to handle the business end, acquire gear, and maintain it so that she can science as hard as possible.
Most of our lab gear is 80s and 90s-vintage gear we've fixed up. You'd be amazed what a little analogue electronics knowledge can do here; people will throw out a device rather than replace the belts or motor brushes. We'll get used and broken gear at a fraction of of the cost of new and certified refurb machinery.
I'm currently modding our venerable Labline orbital shaker incubator. Over the course to two months the tachometer and both of the remote bulb thermostats on it went out, so I'm installing a Johnson Controls A421 to manage the heater and adding a $20 Hall effect tachometer to replace the analog tach. The mechanicals are solid, though (got it for $300 plus freight!) and getting another would be a slow expensive crap shoot.
BTW, readers, do you think that there is an audience for a blog that talks about the details of rebuilding old equipment like this? I'm not big on video, so it would mostly be long form writing with lots of images.
But if your goal is to eventually scale up and build a production lab, you will need to switch your mindset when you get some funding and buy some decent equipment. It can make a big difference from my experience. My company started out at a sub-par lab space with second hand equipment. When we moved to a very well equipped lab, we found ourselves exponentially more productive. Mainly because we no longer have to constantly second-guessing our results due to sub-optimal equipment/reagents. Re-running experiments because of potential equipment failure adds up in terms of resources and psychological stress.
(...)
> BTW, readers, do you think that there is an audience for a blog that talks about the details of rebuilding old equipment like this? I'm not big on video, so it would mostly be long form writing with lots of images
I see a business spin: sell shovels for the upcoming biotech boom. Get equipment on the cheap, fix it, reseller it with a warranty.
This is really inspiring. Contrast with the typical story like Sana Bio which burns tons of R&D cash. My question is: how do you systematically search the potential solution space which may not be your initial area of expertise? How did the solution to focus on easily billable pre-natal diagnostics present itself to a team with oncology expertise?
Make the most of what you have to work with.
I like to build laboratories using research equipment for money-making purposes, not biotech but industrial chemicals.
An ideal way to start can be with a procedure which the laboratory has recognized expertise in, often with a firm foundation of unique or advanced apparatus.
A lot of work involves instruments and they can be expensive and supported by very versatile (costly) infrastructure in academic or industrial research environments.
A worthwhile operation, well rehearsed on an earlier model of the same equipment, can often be invoiced repeatedly in a more businesslike way if the gear can be maintained without completely relying on costly manufacturers' service and supply offerings.
Especially once it's paid for, then even if the invoiceable work drops off you can leave the fully depreciated asset in the lab as long as you want anyway.
Without the institutional overhead a good niche is to offer very limited variety of functions which actually cost the researchers more to have accomplished in their own large laboratories. Plus if you can help them overcome delays or previously unsurmountable obstacles they will want you to work overtime.
If everything turns out good each project pays for itself and always performs profitably after that whenever needed.
But in between your invoiceable sessions you end up with way more instrument time than any institutional researcher, and have more freedom to do whatever you want most of the time.
That's a huge multiplier on the equipment productivity alone in case you do need to pull ahead in the experimentation & discovery department, you can often outperform a PhD who doesn't have nearly as much chance for bench discoveries, and many do not spend very much time at the bench anyway.
Simply too many bright PhDs of all ages would experiment more if they could but there has never been enough time across the board.
With a bootstrapped high-productivity structure, if you disclose the full one percent of your actual progress people will think you're highly accomplished and you can publish it yourself with handouts at conferences.
Until you decide to monetize any breakthroughs or milestones, you can survive or even thrive on the output of your documentation process if you can just issue a single page on average per day of laboratory data to a client. I'm sure there are a number of niches where the going rate is now over $1000 per page and it can be worth working up to it even if it takes a few years. Also took me a while to get a basic report onto a single page for each invoiceable process. After that I never had to publish anything I didn't want to unless somebody was paying for every page.
Plus then you can carry on your research however long it takes.
Without having to do any excess paperwork seeking grants or publishing lack-of-progress reports. You'll have to deal with business paperwork and client interaction instead. That can be just as frustrating but at least it pays the bills at the same time it can build shareholder value.
You don't need citations if you can give clients their money's worth and have an invoice printer attached to your instruments like they don't have at the university or places like DuPont research.
Plus once you're in business you'll have backups they don't have either.
I'll be doing it again, would recommend it.
These things cost time and money. It’s not something you can do in one year. You need five to seven years... if you’re on the right track.
Also you cannot know in advance if a clinical product can be repurposed because that’s not evident till you begin studying your first target (where the data might show other unpredicted positive outcomes elsewhere.)
This sounds like a great example of them focusing on one thing and doing it well. It must have been a tightrope walk not to give too much information away about their IP and get crushed by competitors.
Patent fees for an individual/small entity are only a couple thousand dollars in the US.
That's still quite a budget. Not the sort of thing accessible to many people. What if you're a grad student that just finished their phd (coming off of making 40k a year for half a decade), and not, say, a professor that has connections? Where do you get this initial seed?
While I appreciate that self-promotion is a valuable skill, I believed (possibly flawed) understanding? narrative? that part of the reason why YC / tech incubation / tech can be effective is that it provides a structure where folks who are highly technically minded and historically preferred to spend time leveling up on that can be given easy access to networks with a builtin trust biased to some relatively technical merit, and resources to help level up on networking at all stages (whether that's PG's imprimatur that gives you access to VC offices for late stage companies or resources to help you find networkers in early stage companies).
If YC doesn't acknowledge the structural barriers that make identifying technically meritorious biotech founders (vs biotech founders that have focused on buzzwords and networking) difficult then it's going to have a rough time meeting the success expectation it has for software, on biotech. It's nice to say you want biotech portfolio, but the proof is in the pudding what are you actually doing differently to get it done?
I have not yet worked in a lab setting or where my PI is strict about hours, but from my software background a lot of "butt in seat" time is not really productive. My instinct would be to automate as much as possible and spend less time at the bench. But again, this is theoretical at this point.
If you're in grad school, my strongest recommendation is to optimize first for a PI that will network for you. That is generally the most valuable form of networking, because 3rd party validation generally carries a strong amount of weight by human nature. Plus it potentially takes minimal effort on your part.