Synlogic: Accelerating Synthetic Biology with Fully Unified Informatics



okay thanks for the intro and thanks for everyone for being here today so the throughout my throughout my career in the in the world of process development one of the biggest challenges was basically keeping up processing the data as fast or faster than we were creating it this is often a challenge in this space due to the equipment we use for fermentation processes and through my various roles I've dealt with in a number of different ways but I think that the way that we've dealt with its in logic has been one of the most effective due primarily to how we've been dealing with it with benchley next but before I go into that I think I'll just give a background on what we're doing it's in logic so it's in logic we're creating a novel class of living medicines is what we like to say and we are creating agent synthetically Genet a genetically synthetic non-pathogenic bacteria that's naturally found in the human gut biome we genetically alter the bacteria to perform specific functions within the microbiome without the need to change that microbiota completely and some of those processes conclude such stuff like metabolic conversions or therapeutic synthesis within the microbiome itself treating the patients from that spot and of course the big advantage of this is that we can leverage the safety of probiotics that you'd find on regular store shelves while also getting the advantage of that therapeutic effect for for patients so what I do is in logic is actually work in the platform development group once the strands been engineered it's passed on to us in this group we grow the strain out at much larger volumes while maintaining the same potency that's used in in the patient before it's passed to the manufacturer the patient when I first joined sin logic a little over a year ago the majority of the work we were doing was based around a bench table by a rector like you see here we were using this majority for process scale-up experiments as well as material production that we were passing along to other groups within the company for various types of testing well the majority of sin logic uses benchlink primarily as an ELN due to the amount of data we produce with this type of system we were running into some issues with getting our data up with just the ELN utilization and so I'm going to a little bit about the kind of data we were producing this is this is our average type of been shot bioreactor we were using we had four of these when I joined and we produced two main types of data with these the first we referred to as online data this is anything that was produced within the software of our bioreactors dissolved oxygen temperature pH pH control agitation and so on and you can see a small example of some of the raw data here as well as the traces that we would be looking at within the bioreactor itself the other type of data we're producing is known as we've heard is offline data so this is anything created by the engineers or scientists who are running assays either during or after the fermentation process so that includes optical density readings which is how we track our cell growth metabolite analysis in our broth or you know for any process deviation notes and and so on as well as cell health and potency assays following the fermentation itself so the offline data wasn't a huge issue to deal with it was often times much less data than the online data which if you're if you're considering how many different traces we have here we're running in one of these by our actors for about eight hours and producing a data point for every single trace every 60 seconds or so so this averages out to be either you know but somewhere between 2,000 and 3,000 points of data per reactor of which we were running eight a week so obviously how did what we had a boatload of data to deal with and we went and asked mentioning what might be a kind of a good way to move forward so the first thing they suggested it was to start using workflows which seems to be kind of a common theme here today so they helped us create our own customized workflow with individual steps from the beginning of our fermentation to the end where not only could we organize the process from beginning to end and keep track of what we were doing but also capture the data at each step which was then also moved directly into the data warehouse and you can see here a small example of one of those workflows we've got two fermentations here with you know our experimental setup step followed by the actual fermentation step itself and then our raw data collection and and processing with you know conclusions at the end so each step of the workflow is actually in itself just its own eland allowing us to review and lock those steps on their own so we could meet those GDP requirements that we had as well as capturing it into the data warehouse which allowed us we could you know link any fermentation itself where the data associated with it within an ELN so we could just send a note to somebody else in another group who wanted to know some of the data we had just gotten and they could just you know go straight into our bench Ling Yellen and just find that data immediately this kind of bypassed a lot of time spent tracking people down or you know sending emails or anything like that so it was already saving us a bunch of time but this was used primarily for our offline data like I said there was a lot less of that to deal with and that really fit well with the with the workflow method but when you did a way to also deal with our online data that was coming out of the reactors and so well that was possible to do manually it our reactors were spitting out a an SQL database file so we had to extract each trend from that file rearrange them into a single data set and then use the bench lining web UI interface to manually upload that into the data warehouse itself where it could be linked with our with our workflow data but this was you know fairly time-consuming especially when consider how many reactors we were running per week this would get out of hand quickly and we'd kind of fall behind and then we were spending a lot more time organizing our data than we were actually looking at it which makes experimental planning a bit a bit difficult so what we wanted was basically a one-click solution so after we finished a workflow we wanted to basically get the data from here directly into the warehouse with a really quick method and eventually suggested we look into their API and but unfortunately we didn't have any programming skills within the company so I reached out to somebody that I had worked with in the past named Kai I believe is here today if you want to yeah there is and so Kai as a data analytics specialist and he basically worked with us and created a customized program converter so what this did was we could drag those SQL files directly from our reactor software into a localized network folder the converter would pick it up automatically extract the trends rearrange everything reformat and then go directly up to the data warehouse as well as into a different folder where we could pull that that data out on our own if we needed to and after after a bit of testing and work we got it working great but then it was sitting in the warehouse of course this is a huge amount of data you can't just pull up into an ELN and so when you consider the process development making a process change can often be you know 10 to 20 fermentation processes over you know one or two months time our previous method of dealing with this was we would have everything kind of spread out between either Excel documents or with invention Ling and we extract all this and combine it and then you know move on to the analysis and reporting steps where we could make those choices to move forward with whether or not we needed more work on a process or whether we kind of met our goals this often came you know it was cumbersome it was it was hard to search all data especially when you've forgotten what you've done which when you're running at 8 a week it's pretty easy to do and it made it difficult to kind of keep track of the big picture of the directions we were moving with our research so so just like Michael had talked about he's say he mentioned tableau we're actually using something different called power bi and Kai also helped us with this we basically built pre-designed hot-swappable data visualizations and graphs that we could bring data in and out of really quickly this made it possible to you know do anything from tracking our fermentation trends over time of either individual or or multiple fermentations we've tracked our reference material over time to see how well that's keeping up with what we expect we've even used it to pull together our equipment usage by it by company program to see where our resources are actually being used the most and this is just a few small examples of the kind of stuff that we're capable of once we start leveraging that the data warehouse and where everything's living and the best part is when we add new data into it it's immediately available to just populate these graphs again so we got this program running and we thought it was really great and we thought we were done but then once other groups started to kind of get wind of this we they wanted to of course get their data into the same systems not to mention at the same time our company was growing quite quickly just within our department we doubled our number of bank-shot bioreactors we brought in our real-time analytics machine that was spitting out tons of data and we brought on a both an amber 15 and amber 250 which are 24 reactor high throughput disposable by reactor systems so add all this together and we were had basically increased the number of reactors we needed data from by an order of magnitude add into this that you know we had other groups that wanted to deal with formulation data or downstream processing bio analytics and even in vivo data of testing of our material so that kind of brings us to where we are today working with CAI again we've redesigned and streamlined that converter into something that's a lot more flexible and even more capable than it was in the first place well we were using it for SQL files originally we can now pretty much add any type of data file that comes out of equipment through this converter and then go straight up to the data warehouse that includes CSV files or you know just Excel files or even just text files the converter can be programmed to do pretty much anything we want it to do in terms of rearrangement and processing before it goes into our you know pre-made schema within the warehouse itself so we've already added a number of these items and are still kind of working today on getting those those up and running and that's oh and the you know I think the thing that I wanted to really strike with this is that the amount of time everything that I've talked about today we did in the last year which we looked at a number of different Lin systems that that we could leverage to do some of this stuff and I don't think any of that would have been possible in that timeframe bench Ling gives us the ability to you know user customize all of the the warehouse connections we can make custom workflows which are easy to change and we're even using the request systems as well to to run fermentations for different groups and produce material while keeping track of it at the same time and then with the ELN integration of the warehouse data as well as data parsing using you know informatics programs such as power bi through the api this has really put the data at our fingertips whereas before it was kind of so spread out and difficult to follow and of course the benchlink team was was extremely helpful throughout this entire process and was kind of paramount in getting all this implemented in such a short period of time so I think that's that's it I wanted to thank you know pip reader Lauren Shields was i Benchley rep kai who's here today and then also the same logic PDMS upstream team who did all of the testing of the program for us which was I'm sure a bit taxing hi Ryan again um I have question about specifically for the offline data what's the so you get the text files you do some ETL and you jump into the data warehouse what's the benefit of using the bench laying specific data warehouse as opposed to just a vanilla Postgres database in Amazon the workflow was a big part of that so allowing us to kind of organize the steps that we took you know from setup to to the fermentation run to the kind of after data processing we could kind of have steps listed in there with the ability to also put the data in as well so it kind of worked really well together as this you know getting things where they needed to be where oftentimes we would miss some of that so it just created a workflow that we could not for it not miss anything Thanks okay question is so I'm just looking at the Internet and obviously sample management is also a big deal in studies like that and you know sometimes you have generated samples but linking it to a like specific step or you know specific run we're also in the world more in the process development kind of area where if I'm from SCC biologics is there a way to like organize also link samples to the data so because I know that there is a storage sample storage system that's linked and benchlink – right yeah do you guys use that we are using that we're using I think the refer to them as batches so we create an entity within the the warehouse and then everything under the entity is how we kind of track it if you see data can you link it to the sample and then you can find it where it is in the freezer yes we haven't done that yet okay and we do have our own kind of internal freezer management system and we do use the the IDs that are inside of benchlink to track those samples but we do create you know we call it a pole anything we take out of the reactor which will then go through some sort of assay has a unique ID that lives with invention itself okay thank you okay [Applause]

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *