[00:00:05]
>> We have a started with our opening remarks and so hopefully this is up to set the stage for today's conversation we're going to have 2 main sessions within the supposin for today and as you just heard we do have 4 different scales at which we are considering research into the latest So pandemic prediction and so our 1st lesson is going to focus on the big picture the end to end scale and this session is going to be moderated by about the process and Dr Whittington during the session will have a number of Keynote op and supplementary presentations as well as panel q. and a discussion later there will be a breakout session on this particular topic how many of you have received individual links to the email to be able to participate in those breakout sessions.

[00:01:00]
After the break and there was some group discussion as these breakout sessions will be happening in smaller parallel groups having smaller conversations will then come together in the report out to get a better feel for the conversations that's legs after this there will be a break for lunch if you are only participating in the web in our him out in the breakout sessions and then after our panel q. and a you will be able to rejoin us again here in the web in our after the lunch break following session one to have a discussion on our 2nd.

[00:01:35]
Looking at research needs at the molecular level and we have a very similar flow our keynotes our follow up presentations the penalty Renee and then the break up discussions which will again be happening using a different platform. Those of us who are participating will have already received those links.

[00:01:55]
So after the report notes we will come back together again here in this room women are for some close in discussion and so we look forward to all of your participation today during these 2 different discussions and again if you have questions at any point please provide those questions in the chap box because they will inform the discussion that we have during the q. and a so we are going to begin now with our plenary talks and so again our 1st session is at the end to end a level and so at this point what I'd like to do is turn agenda over to where a session sharers Dr percussion then Dr Whittington who will introduce our plan plenary presenters and lead us through our panel discussions.

[00:02:57]
Thanks a lot of material. Still for those who are us joining us just also really starting with the end to end theme of this workshop so we have a set of really great speakers who will be giving ducks and then there will be a sharp panel Q Many a very rare people come the audience to answer their questions which are so as our next idea Terry speaker I'm really delighted to introduce Professor Martin Murat a from the University of Missouri a.

[00:03:26]
Model is an endowed distinguished professor in by complexity director of the truck science assisted science and advanced computing division the bio complexity institute an initiator and a tenured professor of computer science at the University of materia Dr Murat is a passionate advocate and practitioner of trance disciplinary stick signs during his 25 year professional career he has established and lead a number of large cross disciplinary projects and groups is areas of expertise are network science the i h b c competition to be ology biology current social systems and analytics so I won't go into details of all the awards matter has.

[00:04:06]
One atrocious long career but I would just like to highlight that you can see that he's the founder of almost all our societies you can pick up so you're really a partner to help him to help us get off to a strong start that would be a mother thank you thank you so much.

[00:04:24]
Made here my screen to do that a little shock trauma but a preconceived idea that. Again is it not right yes thank you thank you very much a little it's a pleasure an honor to speak at this meeting and 2nd of the 4 meetings that describe and I'm really grateful for being given the opportunity to talk here just before I start I want to point out the data.

[00:05:07]
Which allows ition that n.s.f. just funded corporate path of putting the link here and the goal is to really get folks interests and gender topic come together and find common collaborative points of interest that we can advise dystopic really or. If we thank our team members the bio complexity Institute our collaborators with us universities.

[00:05:40]
Water for working with this very closely this work is certainly not my work alone but in fact my Port Fourchon is relatively small and despite of time others have done some amazing work in putting this program together I also want to thank videos funding agencies. You know including the o.t. and I c d c but.

[00:06:02]
But importantly and n.s.f. who has been very kind of providing funds to us in the carrying out the work we have done so I'm going to break down my talk into 3 parts the 1st part will focus on the don'ts of Marta's and decision making and I will conclude that part by providing some examples of successful application of markets during the cold 19 outbreak.

[00:06:31]
In the 2nd part of my talk I will outline for the challenges that we have was faced while we support. Epidemiologists in the field. And many of you have all the talk about these aspects I might be briefed in this particular part of my top and finally I'll describe the work that our group has done along with our collaborators with the last year in terms of planning and responding to the call it 19 or break and then I'll conclude with final parts and take other messages that let me let me start with.

[00:07:10]
Why I would like to model in the 1st place and the view that I will have is that martyrs have to go beyond just being useful for predictions in the in time models that support computer actually putting other g.b. should be able to synthesize available data that should be we're able to provide a range of interpretations incoming measurements evaluate a range of options and outcomes the media's response at the G.'s monitor the effect of these this one started us coordinate among various stakeholders and finally they should be made useful not just by computing experts by but by analysts and this this aspect of marketing really goes beyond prediction and hopefully this will become clear throughout my thought as I thought what and 2 and modeling and and that connection here.

[00:08:05]
So from a decisionmaking standpoint. You know we have you are individuals as agents of course and they follow the sense and debris deliberate act loop that I've shown here in the in the simple picture but as you start using and thinking about markets into the n. word such as Right now the key is of course at 19 you realize that you have multiple stakeholders multiple types of decision makers you have citizens they dissidents doing good to day to day job a target these local state and federal authorities and then underground responder just to name a few of these classes and each of them writes and makes decisions by either building a mentor a martyr or using a computer mathematical model in the fall.

[00:08:56]
That is successful markets might might not be compatible in a typical form that we're so used to but both lead to provide context specific information and decision making capability speech of these decision makers and this new is really important as the study and build markets for decision making to support epidemic science and in fact I would argue for other complex systems as well.

[00:09:22]
So let me give you a few examples of successful modeling efforts the other folks in the community in the last one year or so there's a very simple cartoon of how the colored 19 outbreak evolved in time and I've broken down time just for the sake of simplification into 4 phases the 1st phase that starts end of February or so because under June 2nd phase was from June to September the 3rd phase goes from September to January and then the other late night in the 4th place and I the thought is that each of the faces actually allowed us to study different questions different questions by were posed by policy makers into the citizens 1st responders and the Martis were used as a result in different forms so in phase one for instance and this is a incomplete list without a doubt right but hopefully you get a sense for how markets get used and what works at this did or the spirit of times in the 1st place there was a clearly a question that was pertaining to discuss importation of disease from China into other parts of the word There's questions related thing disease better meters and we hear a little bit about it today as well and we wanted to understand the impacts of social distancing and martyrs from fear college I Jimmy not Eastern and Columbia did an amazing job in my opinion on the whole right answer these questions.

[00:10:52]
In the 2nd phase we start up in our limited into a new set of questions along with the in question is of course questions related to discovery says as we have seen the pandemic really come down in the u.s. and many parts of the world Europe included who are going to stand as we open up to society what sort of resurgence possibilities would we have what is the possibility of using contact racing apps and again here work by you mass and others as a part of the some.

[00:11:24]
The use of some surveys that were set up by c.n.n. it looked the those allegations were that you knew the Austin did and testing were that the Africans do the mathematics did lead a very important road in the 3rd piece we started to see the Dimmick you know having a resurgence we started to understand the effect of closing schools colleges are opening them and how we would who hadn't opened them understand and predict the use of medical resources.

[00:11:56]
As well as I understand the effect of seasonality the work done by Oxford Group the work done by Principal Brian Grenfell who's going to give a talk tomorrow I believe and the work done by my colleague in ascorbic a Stanford on did some amazing work again in this area supporting policymakers as they went on and finding in the 4th phase which is when we are trying to understand.

[00:12:19]
The divisions. In the wireless itself and trying to understand the effect of vaccine or not and one of my scripts and Colorado and Harvard stands out in this in this pontiff. So let me move on to the challenges in the 2nd part of my talk. But before I do that let me point out that while we had a lot of success we also felt probably shock in some ways we were not able to forecast the disease dynamics that accurately in many cases although we did quite well we were somewhat surprised to the low usage of digital contact raising apps and we learned that interventions often do not work even though they're meant with good intentions because of lack of compliance and **** into misinformation so these challenges continue to course be in front of us and hopefully in the next round we'll be able to say little bit more about it so let me want to the set of challenges that we face of I'm an end to end perspective and I've broken it down into about 4 different lots of challenges the 1st one which was covered quite well in the 1st workshop that that was that was carried out last week that talked about Sanaa says talked about level of illusion about logical response and the main main theme that I want to present the part of this light as a design is that a single model is not going to help us solve this and to a problem we need to find ways to couple moderns a virus of illusion the human immune system to help him expand and so should nomic systems there's another aspect of end to end that is worth noting which is cases spatial scale that go from a household all the way to a country and to the bloke there's a temporal scale that works for men it's an even smaller timescales immerse in the divide us and and the society was.

[00:14:24]
That is a virtual scale in terms of interactions between between members of the household the community with the county the state the country and to the Globe and finally there is issues after that it processes process of the din host disease transmission individual behavior change community response global response car nation and so on so forth here as a result I would argue that multiskilled multi-tiered multi level network representations and modest absolutely the to the 3rd part of end to end modeling and informatics has to do it right and to take data and converting them into decisions and communicating these these ideas effectively to policy makers.

[00:15:12]
The figure is maybe too crowded but the basic idea that I would like to. Convey is that one gets a large variety of data from different data sources issues about IP issues about privacy issues about commercial use issues about computing challenge the all of them play a big role in access to this data set the markets need to work closely with the data sets that are available produce meaningful insights and then provide them to the decision makers as they make decisions using the data sets and the last topic I would like to cover which always gets asked the model is one of the validation and uncertainty quantification in the end all of these models as we all know from the famous courts are wrong but I think they are useful in many cases and I think if you want to make them artists for epidemic size to be useful I would argue that notion of validity has to be extended of a from the standard notion of retrospective and predictive energy.

[00:16:16]
And you talk about internal structure validity as well as understanding the role of markets Lucy then certain assumptions are failing to capture here who are dynamics and then adapting them that other challenging questions are there to governance it takes logistics diplomacy and one have that all important but I want covered or can today stop so with that.

[00:16:40]
The time I have let me try and quickly outline the program that we have built here and that we used as a vehicle to support the d.s. authorities during the last one year as we as we work closely with them as they responded to the covered $99.00 break.

[00:17:00]
So the way we approach this problem is this concept of nettle deputy neology. We did this work early on it back in 2005 in our major paper where we showed how that works can include a present. In Silicon social contacts and how those networks can then be used to drive into important policy insights we have taken the program quite a bit ahead where now we're able to build synthesize these networks using very detailed data from digital basis from surveys and such and build their realistic social contact networks we have just completed the 1st version of the social contract network for the entire us and we're working towards improving it as we go along the 2nd step we use a social contact network in fact it's a multi-state theory multi-skilled network it is about aspects of disease transmission it is that representation of Indonesian collective behaviors we use the martyr's Ryan simulate Lidia's contagions process and some of these networks to study how busy is my propagate of information my proposed gate what sort of actions might be taken and then use all these and information sets and results that the simulations produce carry our country factor analysis the study strategies from medical cost protections and to do analysis of what might happen under certain Somalians so that nothing you have about 5 minutes left.

[00:18:33]
And I think I'd say that should be enough thank you very much. So let me give you a practical perspective the work we have done weekly because I think this is a very very satisfying aspect of the work we have done for the last one year we have been supporting multiple agencies and groups out there we actively as that is sponsor pandemics I want to really acknowledge the work that Brian Lewis and Julie have done in trying to coordinate this on behalf of our group.

[00:19:07]
$8080.00 a part of come together to try and do this week after week and we're probably in the 52nd week right now of presentations to various agencies from deity to media sparks of the common that the Virginia where our local agencies. This is been a very. Satisfying effort.

[00:19:29]
We have learned how to ride to the ultimate But I'm excited as we support important policy questions what sort of updates and be fix do we give we said we give what 6 to 7 briefings of eek and if you go to the United States with your Department of Health that page you will see we just report that they put out based on the margins and data that we produce things which give things like situation assessment literature survey marker projections negative summaries and such similar work is carried out for other agencies but I'm going to focus on the work you've done for our state but see if you do the rate one particular piece but instance every week we try and do projections we try and study where the pandemic might go under different scenarios and this notion of sin that scenario analysis and projections has become more and more important than just doing predictions.

[00:20:35]
We have built out the ideas dashboards into those we built out the surveillance dashboard very early on Johns Hopkins a very nice dashboard as well done around the same time to build out other dashboard for national resource demand modeling for medical resources with the group at Stanford led by Michael Bernstein we try to collect using cults those methods medias interventions and policy that reported by counties in the u.s. and they recently wrote that you know let's go next crop to build a mobility impact dashboard based on a very important paper that he and his colleagues about in nature and this dashboard or all the news release groups across the federal and state government we have also supported the c.d.c..

[00:21:25]
Well with 19 forecasting ensemble or we've been providing information to them every week for the last about 6 months or so of that they have been very active has done very well in that in that program but I want to again acknowledge Nick likes group and other. Forecasting methods especially letters from Los Alamos next to CMU and others who have participated in this to provide very meaningful forecasts in the various agencies that look at it we have built a base in on some that ourselves and what we have done is we have been able to build a multi scale system where we can forecast the disease outcomes for the whole country.

[00:22:08]
For the state of Virginia and then every county in the country so here for example I show the whole u.s. outcome we have cannot compare what you know but we have the same basic map for every state in the country and then every county and in the country for 3500 out these are are being studied in a modest on a weekly basis if you want these models to study a number of country factor analysis questions one that I want to quickly outline is where we used to pitch books at the computing center and times when a 7 page book supercomputing central providing this it's p.c. access to us we are able to compute and estimate the past of though it in terms of medical expenses and the paper that talks about the bigger economic cost is also under review and the interesting part here is that we have to really build h.b.c. grid their competition was done in the local machines so aspects of computation was done nightly in the pix multiple computing center and then the data was brought back plus supercomputers here please insights from it and one of the insights that we have to report in our nature scientific reports paper is the medical costs which we estimate to be plenty of dollars within the worse case but we show that interventions can actually lead to lower and lower medical costs this is not surprising of course but the numbers are not to meet other important interesting.

[00:23:38]
I've concluded my talk in terms of results by this made recent exciting result that we have gotten this papers and made archive you can do it because look at it what we did was we used the digital pin that we had built up to study vaccine privatization strategies and that results were quite quite exciting and interesting from our perspective we showed that next if you vaccinate individuals based on their contact time or degree in the social contact network you could actually end up just in the state of Virginia and just in 2 months.

[00:24:15]
If the now than 200000 you were infections 702-1000 fewer deaths and 8 to 13 percent fewer hospitalizations and if you extrapolate as desired results across us we think this would have a significant impact because in this is a really important strategy to consider especially in resource poor countries then I do not think vaccines are going to be available anytime soon or and they're available they're going to be available in very small amounts you know the idea of using high degree nodes to x. in it is a folklore at some level what is really new here is that nobody has ever studied this when the network is actually changing in time because of the pharmaceutical interventions because the vaccines being available not at right at the start but vaccines being available to a schedule and all the behavior that patients individuals are doing as the epidemic progresses this the fact that this result hordes even doing this bomb makes makes a problem very interesting for us.

[00:25:18]
So let me continue the lessons learned for from our work that we have done. This includes a couple of different lessons one responded working closely with stakeholders and policy makers really helps us make models that are useful explainable transparent and allows us to build better models we found out that we need to be a child and flexible.

[00:25:43]
I cannot overstate the importance of. The plant discipline team size this work would not have been done had we not a similar theme of this size and with so much expertise across the board and finally I think communicating scientific results in such situations need to very thoughtful and deliberate So let me conclude my talk by saying that an effective strategy to reduce the global pandemic must detect the timing and location of occurrence anticipate public reactions and then of actionable interventions that enable targeted effective response I think this is not the 1st pandemic we're going to see in the next few years next I hope so hopefully not we won't see many of them in our lifetimes for sure but the looming challenges from climate change and the microbial resistance synthetic pathogens input that makes urbanization and global transactions and expectations of finally information from everyone make addressing challenges that we might face when the next epidemic comes which certainly would happen to be would be something that we need to be prepared about so that I will stop and hopefully we can have a question answer session at the end of the teacups Thank you thanks a lot Michael for that great.

[00:27:13]
Cockerel So I think. I land over there by 2 now microorganism crystal for the next next. All right I'd like to introduce our next speaker. Dr Devore Peterson. That appears in the holidays for that u.s. Department of Agriculture. And now the principal investigator for that you're not obese and long term ecological research program of Cruces New Mexico Dr Pierce has greatly contributed to the Prodigy search enterprises senior advisor to the chief scientist at u.s.d.a. and as a member of the National ecological after observatory network Sprint directors in addition to be amazing array of service to her fast and to her profession she is maintaining outstanding on Liberty search productivity and landscape cross scale ecosystem ecology research tracker led to the early action as a fellow let us say in the American Association the answer to that and like to pass this over to Dr Peters thank you for joining us today right thank you we see if I get my side story here.

[00:28:41]
Ok so what I'd like to do is just provide or if you're one of the projects that I've been working on for about just what the last 5 years some of what is called Just with this is a problem that was presented to me I was at a actually at a workshop and I saw a person present this and then I sort of like skiing with him later on how would you actually solve this problem so it is going to be a disease it's an animal disease from early in the u.s. it affects horses because it can be transferred to humans but it's primarily a disease that affects horses so our goal is to develop a strategy to frame were for complex problems that's what I work on you know you guys are drawn into this is these problem because it was something that heiresses really interested he was so are pro-choice is a spatial temporal modeling across killer actions coupled with human go and learning and they were just applying this to visit you are so Titus it's infectious disease of livestock as I said that is causes economic loss Cortis restrictions initial international trade.

[00:29:40]
So it as a mole disease it actually works quite well because we had access to the data through another u.s.d.a. agency you know the Animal Plant Health Inspection Service it's reportable disease it looks like foot and mouth disease in cattle and pigs which has been eradicated from the u.s. since the 1920 s. So it's a big deal right to actually get this correct so they actually report it to a focus if this goes out and check to make sure it's not quick mouth disease but then they're actually track of the essence of d.s.p. there are multiple hosts it is that the actors in the u.s. is primarily horses comes from up from Mexico does that it's endemic region that's where it's mostly in cattle and other multiple insect factors as the brawl breaks occur approximately every 6 to 10 years but it seems like that frequency is increasing and it is the most commonly reported securities of life so I can in America so it is kind of a big deal so when I 1st saw this figure and it showed.

[00:30:42]
This I was at a talk I said I saw this pattern and so it was the person was talking about it worked on vs for a very long time especially in Mexico and he he said we don't know if we don't understand right is the patterns here because sometimes 2004 if the disease these are current state of the points are there some ways it will stop in Colorado Other times it was all the way to Montana he said we don't understand so so this was this was a question Michel Louis was what explains this spatial pattern in d.s. occur and they had a lot of data but they hadn't actually analyzed it in that way.

[00:31:24]
Ok so 1st of all it was seemed to me as I listened to them that we needed the trans disciplinary team so less Dr Luis Rodriguez was the one who's actually been doing this insists that his Ph d. he was overall a just they also working with Avis in Colorado and animal epidemiologists but we needed really was Ok it's a disease that actually you know there's a sector so many people actually understand it's x. So that's a hold on the people in Kansas for that there are people who actually that and that's still doesn't get us into the environment right so we need to the people who connected the disease to the environments like a real scientists so that our people in Fort Collins with the air s just and turner and then as I listen to them and actually some of these is that vectors part of their life cycle in water and it's flowing water so then we need an.

[00:32:12]
Eco hydrologists and so I brought in eco hydrologist and we've now this was our so initial team was very small it was 5 and then we expanded out in our national working in endeavor to Mexico as well to catch the Hanley at New Mexico State University. So 1st thing I asked was where is your system's diagram they didn't actually know what that was so we had to put this together and I so I put the future of your together right listening so I didn't want to do is to have sort of a correlation analysis which is very typical for geographic disease spread as you sort of correlate you know viral drivers with the current state because we need to be able actually to manage this disease understand the processes underlying different actually do predictions and sort of provide it put to management so I wanted to relate all the information about the processes which is you know whether the insect population dynamics or how they spread it to the host or how host then it spreads if he moves or how it's you know goes back to the insects really I feel he understand the processes in order stand you know how to move forward on this and so this is a multi scale diagram of the green area is the local scale sort of environmental conditions at the level of a farm or ranch and then build landscape scale you know we can move animals they can be transport and start infections from place else there are also insects can disperse so I can see that for the latest gave to broader scale you know broad scale climate drivers so this was a diagram no put together a lot of courage to 6 months just basically bugging the people who actually understood the disease quite well and coming up with this figure and this then led us to the data that we needed right but they didn't but again we didn't want to just use the drivers which is like Forest for example soils what will what is it about soils because it was actually complex right so what is it about soil so we went through each of those key processes in the previous figure and those are the ones in the elbow and then with each of the major drivers rights of climate so aisles elevation and then we ask what is there about those so we never have the data that we need so we trail we call eco transfer functions we work through a what do we know you know from the lab you know a number of expert dollars to a water temperature for example and we work our way through until we can actually get to a relationship that has hypothesized relationship between some soils variable here it's mill available water.

[00:34:28]
And then and then the density of eggs and so they have actually collected that but that's not possible size relationship which tells us Ok if 8 of you see is the soils verbal that we need right instead of still you're saying you're clay right so we did this for every one of our processes and all of our driving variables drivers to get to the variables that we need and it would just flip around to go through it again so precipitation what is about precipitation right so this is how we actually came up with the variables from our drivers so soils as a driver beastie is the variable and then we go get the data so we is came up with a very what we consider a data cube which is in time and space and variable named we end up with $484.00 that is with 10 that actually do all the actual time element so it's a fairly large state a cube and then we know we are in the now since and so we only have presence data for at that time for vs So we then after we standardize it harmonize the data and we did it for 4 kilometer by foreclosure Grisel across entire western us because that's where the b.s. occurences So we then after we had all that harmonize you could then you know why it is so because really have presence data were a little bit limited in what kind of analyses that could be done so we used a max in which says it's a species distribution habitat model is basically an elsewhere question across the entire us and then and then that then provides us with a really interesting result so I think we have time to cover some of those results between quickly being so basically a member they basically knew nothing right or they knew a lot about the viral 2 never takes they knew like where the cruisers were they had to meet Dia's about what the vectors were but then they didn't really understand ecology right so their 1st analysis we do is this next at Morley and what really interesting is that we used to.

[00:36:20]
Break the 2004005 the 1st thing I did was separate those to the operator to 2 separate pieces. Incursion you're an expansion you're it turns out that different Orton but you need all of those variables right so you need any claim that you need soils you need elevation horse density of course which was out in the 1st year the incursion the or it's really the one important variables is distance to water Well that's actually black fly habitat that's one of the vectors in that in 2005 when you look at it again it's quite in that it's its force density is distance to water but now the soils property comes up and that was the one that we actually put in there because to a quite easy as another it's by images is another director so what the data is showing is that you know it's one of the actors reported in the 1st year likewise and then.

[00:37:11]
The body image is a separate vector because of Porton The 2nd is so it's only by sort of communal combining these process based information with a broad scale picture of climate and other drivers who actually come up with that analysis result and then we test it right we had a 2nd outbreak in 205-4015 which is a different outbreak and actually had the same result the same variables so 1045 are related you know genetically right at 14 to 15 this actually environment is more similar so 2004 is more similar 10140058 similar 1052 as environmental variables and so it it turns out that it's the virus actually driving this patters and that and not the not the virus.

[00:37:54]
Back to Peter's group just another minute Ok so we move on now to deliberately What is strategies based upon this so turns out that it you know it's more related to things you know in each year the actual specific variables to time and then we apply the same approach to West Nile and Tripoli which is another another disease so I think we're comfortable using this approach I think the general public she works really well but in terms of back to more diseases and they were predicting future operates right so we actually came 200-1020 was another outbreak which you're very surprised at because what we were told was sort of every 6 to 10 years and now is there ever for this or in the middle like like over how we actually predict where you next what's right how are we going to predict this given that you know and so that we also the answer to using a bunch of different specialty or a model is to actually make his predictions 42001020 while it was a curry and it turns out that you know 2001020 years now what you're doing to now said is very different so that we showed up in eastern Missouri.

[00:38:56]
Kansas said never been that far east before so sort of stretching I think that what we just heard you have to be flexible adaptive exactly because you know that these things are changing all the time so now we're looking at sort of why is it come across the border why is that it is different patterns to the different outbreaks and then you know how do you actually then predict used for future so so I think they it's ait's as we put it together transistor team for an elk of all times this is actually really important to recognizing importance or processes how to relate that to the patterns and not just a user pattern Elsie's and then we just have very regular team meetings who are for us understand and be adaptive our challenges are limited ability.

[00:39:34]
Vectors we don't know very much about their habitat we have a lot have prophecies now about their importance they were going on and testing knows this information is being given to simulate to managers the horse owners in terms of what what's going to happen we could have it again this year so we're trying to break a prediction for 2021 and then multiple computational to schools so we start with.

[00:39:55]
Lucian networks process. It that's all I had just sort of are the papers if you're interested in we have a story map on its readers or works through this in more details then there are writers here this is a very slow it's grown is a lot 15 people it was originally 5 now that's all I have great so thanks again to beat us with it great Dhaka So we now want to our next speaker and this session Dr Rhee and Kate from Los and a Los Alamos National Laboratory so his research group focuses on developing mathematical and quantitative purity is and to understand the sped up widest of the wireless muni interactions and wider Aleutian of the dynamics across multiple scales of biological organization.

[00:40:51]
That is interest in the last cellular populations gets since January 2020 has been working on modeling the transmission dynamics of Star School we do across the globe most recently more recently his work as well because trying to characterize ing within hostile in a mix and immune responses to the Star School we do infection before and the n.l. he was a tenure track assistant professor of mathematics at the state and he did a speech here to bury College London followed by Paul struck at u.c.l.a. And so let's welcome our Nick speaker going.

[00:41:23]
Over to thank you for the introduction and thanks for the committee to present our work. Let me share my screen name. So I'm going to talk about as the meeting keep Ramsar values for no more infectious disease outbreaks and implications for control I'm going to use my work from a source close to him as an example to put our work into broader context I'm showing you the c.d.c. response framework again basically it categorize different stages of response.

[00:42:08]
Prepared to make intervals and the pandemic intervals so the questions on focusing on the really prepared to make intervals where a novel pathogen is identified and human to human transmission is reported during their stage to the key question is to assess. The potential of the novel operate and find out of the most effective control strategies to control the operate before it becomes a global pandemic so tool particularly important primes have values.

[00:42:42]
For assessing potential are early epidemic exponential growth rate later are here which I will introduce later on and the basic reproductive number are not these 2 primitive areas characterize the end of operate in the absence of intervention 1st so it sets the baseline and therefore essential for forecasting epidemics trajectories evaluating effectiveness of intervention strategies and projecting the herd immunity threshold the last item becomes more and more relevant as to day one when vaccine stars.

[00:43:20]
So let me start with the estimation of our little our exponential growth rate this asked mission starts with data collection and then mathematical model to Tater to estimate the exponent of the exponential function later are here so the difficulty here is really to assess the biases and all certainties in data collected this is particularly relevant for novel pathogen where we probably lack of their nose to tools to confirm a in fact it's a case.

[00:43:55]
And because the. Surveillance intensity reported cases may not penned of picture of the food scale of the epidemic or this is well demonstrated by the way to operate early in 2020 there are other factors to social and economical and political factors are. Involved in data collection process makes the data even more noisy and the biased so the central question I think during a early operate of know what has been is how to evaluate errors uncertainties and real reliability of estimation due to all these uncertainties if you look at.

[00:44:41]
If you look at the estimates of laser are for early to operate. You find a vast difference as to Miss early estimates were made from Groups in Hong Kong and Europe this is suggested the virus grow at a relatively slow rate point want to point $14.00 per day translating the doubling time of 5 to 7 days this quantities are important I want to point out this quantities are very similar to the estimates for SARS Kovi one so for a long time period we thought of this virus behave similarly as SARS and this paper this this works made.

[00:45:21]
Papers and have been highly influential in terms of policy making however some other groups including our group showed that the initial growth rate is actually how much higher the virus is more contagious than SARS one the estimated growth rate is the point $22.00 per day that translates to a doubling time of $2.00 to $4.00 days so just to you know straight the vast difference of the estimates using this a toy example if we assume an epidemic doubling time is 7 days and I assume at time 0 there is a one infected individual after 2 months of transmission we would expect approximately $400.00 infected individuals this is a relatively slow spread as we've seen for source of a one however if the doubling time to 3 days we would predict after is through Master transmission there would be a 1000000 infected individuals.

[00:46:21]
So actually the latter case is what we experienced in Lombardy in Eataly or in New York in the u.s. for example so given the different asthma so how can we be sure or as soon as our really reliable and which ones are more reliable than others so for the past year the 2 lessons we learned were 1st cross validation using multiple data sets.

[00:46:48]
Each data says it has its own biases and certainties so what we think is if multiple data standpoint you're the same conclusion that increases the estimation of conclusion. Reliability of a conclusion so that's one important aspect to consider given the high. Uncertainty in the dataset we crack steering initial operates 2nd is predict whether a model estimation and a conclusion.

[00:47:21]
Our consistence visit future operates so as a mention to be asked me the initial growth rate should be very high higher than point 2 per day so our estimation are different from other groups so we quite real hard to make sure we believe our own estimates so we find out today says that a b. did in the years for inference the 1st dataset was the curve by date of some to master published by trying to c.d.c. if you fit an exponential curve to the reported case this is by symptom assets you realize the growth rate is higher than point 2 per day another piece of evidence is the curve by a task constant Derian later generation to 20 from China if you fit across curve the growth in test count also.

[00:48:10]
Much higher than point 2 per day so both state crossable of this or conclusion that the wires is highly contagious is very different from South Korea one predictability as the viruses spread across the globe to be watched closely and estimated the early growth rate in Europe and the us we find that although there have fission ages your estimates but in general they estimate that the growth rate is higher than point one nonparty again these as mission are consistent with the as many as we made initially from one part in particular for the us the closer it is actually very high this was a most later from the from New York.

[00:48:53]
So from the as the mission of a little r. the x. exponential growth rate and doubling time we can ask so many are not valid if we assume a mean syringe of a 6 or 8 is for the 2nd time there are some complications into the syringe about to use for now for a 2nd time or just happy to discuss later on but I'll just say that a similar interval we think is correcting to what you used to estimate are not the estimates or are not occur from across Europe and the us that there are not where is significantly higher than widely reported as are not value of free there are not where it is likely.

[00:49:37]
For Tuesday 6 which suggests the herd immunity strain but must be very high maybe higher than 7 by 80 percent so from the estimation of these 2 primes or values the question before the op become a pandemic is really what its implications for control and how best to be can control the virus before it becomes a pandemic so in early February just going to just a minute run Ok sure.

[00:50:09]
So this is. The last resort I'm going to show world b.. What will be realized that the conclusion we made based on the as a nation of a high growth rate high value of our not the possibility of say is symptomatic and present to matter transmission we estimated that tast traits in accounting alone wouldn't be able to launch with a virus even under the most optimistic scenarios so early strong socialist distancing efforts are really needed to reduce the overall contacts and stop the virus so this pollution hasn't been well received in the early days of the break.

[00:50:55]
At that time most people believe that if this is this wasn't misread to the rest of the world that we can control the virus as a. Way to but data later shows that or estimation and conclusion are reliable so just to wrap it up lessons to be learned for a novel pathogen operate and how to accurately assessing at a make a potential 1st data probably would be noisy and biased and the estimation might be highly uncertain to base to test the real estimates cross validation or dataset and predictability this truth has our lot are largely absent from early studies of the epidemic which may lead to maybe policy making are made.

[00:51:49]
Based on certain estimates so we can do better next time probably and also I want to mention that there is a difficulty in publishing results that are different from already published. Especially already published research from established groups I think for know what has. We know so most so little about the pathogen and the data highly uncertain it is expected to have reached different conclusions of based on different analyses and it is it is a critical for policy making perspective to have diverse perspectives to for policymakers to make policies based on the best scenarios as well as the worst scenarios so having diverse perspectives is important to realize that I would like to thank the organizers for the symposium and the land.

[00:52:43]
Who made the work possible and also my form of a mentor in Paris a Los Alamos who taught me about constructing credible models and making reliable predictions like to thank founding from Los Alamos as. An age is that my talk and thanks for listening. Thank you so much Dr Kay for your presentation as well been really excited about all of the presentation so far this has been agreed group and I'm thankful to everyone for the questions that have been left in the chat box as well as we have been gathering those questions together in at this time we would like to invite both the plenary speaker as well as the this No 2 speakers to be present to answer the questions that you all have presented so we're going to open up now to our panel discussion and for some q. and a there in the us with about 10 minutes from any.

[00:53:51]
And so Dr what update Dr Peters and Dr Kay if you would mind just making sure that your video is on and I'm getting when you are ready to answer a question and I believe that we may be joined at times by our special moderators as well who maybe just one for the supporters they've been in the human a as well.

[00:54:17]
And so I'm actually going to give the options of the recession moderate instance and open up the panel with the questions if you would like to. Thanks a lot amount of and. Thanks again to all our speakers for really diverse and interesting. Perspectives and experiences in this problem so let me start off by asking one question I guess I do all tree speakers which I think they all alluded to briefly Rich in their talk but it would be great if you're going to elaborate like this is to do about success messages like so how do you measure success in such endeavors at the end of the day.

[00:54:57]
All Tosk I mean I think this is related to the problem of addition but also in other fields having clear objective measures of substance as that going to explosion the government of like she's running or Ai techniques to help fill the gaps right so are the goals and measures Korean in this pandemic Stace.

[00:55:15]
Do we know for example for some of the decisions which people are taking already measure success there yeah I'd throw this open to all the 3 speakers because they all talk about this in some problem so maybe we can start in Monaco and brand. And so that be brief so that others can comment and I think it's a beautiful question and a hard one of course in certain specific areas one might Bieber define at least some early measures that you can you know use.

[00:55:47]
You know we have done the c.b.c. forecasting challenge from our 67 years all of us and at thing certain measures for your predictions for for a steady state disease you know meaning things that happened. And potentially been developed even then we find challenges that every year is different from the other years we can potentially use other metrics of success we should tell us you know things about how the society function better in the end counterfactuals are very hard to measure against my opinion and that's something that I would like to in fact say that.

[00:56:24]
In all of these cases. If he can we can I try and put numbers to the value of being some actions so that a certain event does not happen we would be in very good shape and that's really very hard to do because if you if you did very well and that event did not happen I'll condemned it gets unnoticed and so you have to justify so I stopped there that Deborah and others.

[00:56:55]
Yes we were we were for that we will to do the verification so we had Brit outbreaks so we could actually do or model development in the 2405 test to tell you. That there was are pro-choice is the ultimate the data back which is the classical it waited to do verification and then of course the next.

[00:57:19]
Actually it didn't work right so you know if you change it right but they have to have it we have to acquire that the data actually be countable about these predictions and this is a worry about climate change the climate continues to change you know how do we actually make that moving target we have basically we're using our historic information to make predictions about something that's changing in that that's much more difficult.

[00:57:47]
For Success in your field to. Transit all these saving lives. For and that makes it so hard. Hard thing to measure because you concert replicates the pandemic as you do if in the laboratory so it's hard to evaluate how many lifes one saved. So that's a really hard question I grieve his.

[00:58:13]
Other pen on this but in terms of a model inference I guess success. As what the mice to I swear cross validation and predictability of them although would be in a measure of success that's what I would thank. Thanks for much study. Yeah I had a question for Dr Peter so given what you found and the driver is outbreaks how might climate change in fact.

[00:58:50]
You know that the average that we see in the future. You know it's a great question and so because as I say we use new story information but we're simply not these relationships continue so what we've done is to use are these are called pattern process relationships and so it's we sort of then go back and say well how did would you know change of precipitation effect of inspectors right so doing sort of court missions is sort of getting back to the drivers and process relationships and then and then running and their models again so we're doing those climate change simulations with the other things we're doing is sort of we step back and it ecology a lot of times we sort of graph.

[00:59:31]
A scenario because of their literature and we actually have a climate scientist now on staff and so she's saying you know that's not really what where you should be doing so and so she's actually providing an analysis of you know for this space you know which climate model should we be actually using So we acclimate it to actually say Ok let's actually do the climate serials correctly to put into the models of the so that's where we're at so thanks Dr Peters for that answer so this is a question which is coming from the audience which is I think again not pretty general so maybe it's starting to whoever wants to answer I guess on the planet a how do we ensure the needs of Global Health care workers are considered in all of your research efforts is that is that especially you do need to do some things differently already called us really naturally to promote your.

[01:00:24]
So I can start this certain segment thing to get a good question radio demographic groups are important for the producers medical workers clearly are some of the most important ones and the mic comes to agree the way we have martyrs we have played a big dent into account when we've been a digital to enough recipient of the state we have information we are not perfect but certainly we have information about the kinds of.

[01:00:54]
Activities and professions people might have that allows us to capture certain aspects of it enough course very precise information about individuals is very hard to get and probably we don't even want it for privacy reasons but why invent they can and I think the margins that we have built at least try and account for that demographic report that you know so those so they are actually part of the problem for us right so were ours is a disease that can be spread to humans but it just sort of get sick and get well weeks later so that's a risk that entails so I don't think that really we have to we have had haven't had to address that we do have p i issues with our data which is sort of related to what you're asking that we do have to protect the status field online if this is at the county level and we have access to the actual locations so I think there are sort of those kinds of issues as well that we said address.

[01:02:00]
I guess another question now on to coming from our audiences. That the public awareness of pandemics as high as it is now what do you take me as a research community can do that be couldn't have done in your back to significant to push the on will on the science of the next Again this is really a question Richard like open so maybe we can try to clear.

[01:02:29]
Yeah this is a great question a. Very proper question I've been focusing on very specific problems for a long time still probably my also but I think surveillance is really important area to focus on to to know what's what kind of the buyer says around the animal reservoirs because most of the viruses jumping from animals to humans then and also how the they jump from reservoir to human to understand the human and reservoir interaction the interface I thought that would be a very area that screwed you would help to prevent the next and I make so I can say a few words.

[01:03:22]
That I think islands question is really a point I think the huge awareness right now there's already. Things in play right now there's a directive from the president's office as you saw with James to establish a national carnations Center which is going to really change what. Its shape and form of course we are right now but really going to happen the next 6 months and that that will allow people to come together with one of the things.

[01:03:55]
We were thinking about as a group is to really realize and appreciate money to disagree science because of this this lie low based science is not how this problem is going to get sore and maybe if one thing I did this is even stronger signal for me to continue my work is we have to work in a group with people with very different backgrounds and insights and expertise without that I don't think we can make meaningful progress on any of these problems.

[01:04:27]
You know I would agree completely we we pulled together a team that had never worked before together for doing so and the only way we made the progress was listening very carefully to what I'm going to say respect what everyone has to say and then. Putting that information together into a into a systems of moral rights that had to go somewhere so actually put is together into a conceptual it would agreed on they could see where they fit in how of all the pieces fit together I think that's really important if the other thing is we have a lot of data right we tend to have a lot of data we do we have a lot of technology that the you know but I think that we have to then provide it if that information in a way that people can understand so I think that that's you know other pieces specially Toad's sort of about the public is how do we actually provide this information so that we do have a lot of results and how do they get the right before missions in that the they can actually understand trust.

[01:05:21]
Thank you everybody got my last question I'll fit in far and I think really take a break and this comes from one of our participants How does social network information across the u.s. consider on the ground of contacts especially when it comes to minority communities so. I think this question would be for not.

[01:05:42]
So how do we how do we consider that systematic bias as it relates to our minority communities. I think it's a good question and I was thinking about it when it was posted so 1st of all. I want to college to see the real problem that we have in the current epidemic in fact almost all law school disasters minority community always gets disproportionately.

[01:06:08]
Impacted and there's no this time is no exceptions I'm very unfortunate that this is happening this way. But at least in our modeling say McChrystal. And the person was a question we have tried to again build a digital to win that prize and that counts for the so if you look at our counties and the presentations we actually have neighborhoods that have folks with those demographics they do certain kinds of activity that again deflect the extended up available to surveys and other means activity and they do so in our digital to in this effect at least gets captured structurally and we do see as a result impact of these communities that is somewhat different but I think the question is is it in part one I think as we go on improving our network further any information that can be made available in a manner that is 1st.

[01:07:04]
Speaks to that community but still it's privacy preserving and confidentiality preserving would be very useful to put the markers I don't think the markets are done by any means we really look forward to working with folks who can provide us this information and want to believe a reflector everybody's a sentiment when I extend thanks to all of you or our panelists for your remarks and thank you so much for helping from one this conversation and for the value that you bring based on your experience in your expertise.

[01:07:36]
I'd like to also thank all of a participant who asked questions and who provided questions and chap I know we did not have time to address every one of the questions that you asked but I do want to assure you that all of those questions are being documented and will be incorporated into a report the goods and I set up you can see this workshop so your point of view has been captured and we do appreciate all of you for that so thank you again to all of our all of our panels in our seekers at this time thank you thank you so much.

[01:08:07]
I want to take just a quick moment as we move forward in our agenda we're going to be moving to our breakout sessions next and so I want to just provide some quick just a point of process for those of you who have registered for the breakout sessions you have received a separate e-mail from n.s.f. which includes at different zoom link.

[01:08:29]
For the Xoom meeting which is different from the Web another that you are in now so if you are going to be participating in the breakout sessions Please make note that you will have to exit this zoom webinars and use the alternative link that was provided to me via email an order to open up as a meeting and the breakout sessions will take place then so we have a few minutes over break I do encourage you if you are going to be in that soon even Please make sure that you enter into the meeting before you leave for break that will give us the opportunity to put you into the appropriate breakout So if you are going to be participating in a breakout session Now we asked at this time leave the web in our move to there's a meeting and do so before you take your break for a grab your coffee or water or whatever it is that you need for those folks who are going to be joining us here in the web in our own book will not be participating in the breakout session we will be back here in the web an hour for a wrap up to its end of the day and that'll happen this afternoon at 225 we will continue with the session 2 and our discussion on the molecular level so thank you everyone we're going to break at this time then for those who are in the breakout sessions we look forward to seeing their.