We have like a really nice case study. I don’t know some on the website, but like where we start with like a big, like would we make like 3,000 or 4,000 synthesized enzymes from all over, like our big, like you were mentioning that big 2 billion or whatever genome like gene sequence collection. We go find all these potential genes from our proprietary database, print them, we’re going to print at twist, put them in selves, test them. We do that. And then we train our own machine learning model just for that enzyme, okay? Like that data goes into a model just for this particular enzyme. And then we do it again. We tell the model now to say, hey, give me 3,000 new ones to make. We make them — we do it again. And then the last improvement, which is the 1 that gives the most improvement, we only try 100.
Because at that point, we’d like generated data and trained our model well enough that the model is now becoming increasingly predictive, and it’s saving you the lab work. And what’s really cool is now if the future customer wants that particular class of enzymes, we’re not starting all the way back at the beginning. We now have this like trained up model. So that is — we are selling that today. Now what is the right business model, I think it’s partially what you’re asking, like, maybe I should just offer that as a service. maybe you should access that code base, but you do the lab work at your lab customer X. Today, we don’t think that’s great. We think there’s a real nice alignment between the throughput that we can run on our lab work and the data asset because there’s just a lot of high throughput like both analysis and lab that it’s better together, but you maybe could imagine that in the future.
We are not today actually offering that data asset as a stand-alone product. It is only offered integrated as part of a larger cell program. Does that make sense?
Mark Massaro: Yes, it does. Thank you for that. And if I can ask one last quick one. Your guidance for 100 new programs beat our estimate of 80 for 2023. I’m just curious if you’ve seen any changes in lead time to like close a deal? Like what is your typical lead time from identifying a program and then closing it? And maybe I’d be curious, if you’re seeing any changes to the components of your funnel as we enter 2023 here?
Jason Kelly: Yeah. So I can say two things. One, I’d say it’s the healthiest I’ve seen really since we started Ginkgo. And that’s one of the points I was trying to make earlier is I feel generally good like our ability to — and by the way, this is going to vary quarter-to-quarter right? Like it’s a similar thing like it’s not like total metronome here. But like as I look across the year, we have a big set of potential both new customers that have new programs attached to them, but also increasingly inside sales into our current customers, right? So I was talking about that in biopharma, like you have — you do a manufacturing deal and then next thing you know you’re doing an R&D deal. That is that the thing is starting to shorten our deal cycle — we have a very different, maybe to state the obvious time to close a deal, if it’s one of our current customers as a close deal, I mean adding a new program versus a brand-new customer.
