Biomedical Informatics and Data Science at Yale

Transcript

• 00:00Thank you all for joining today.
• 00:02Today we have Lucia Ohana Machado,
• 00:04who's the Professor of Medicine
• 00:06and Professor of Biomedical
• 00:07Informatics and Data Science.
• 00:09She's also the Deputy Dean for
• 00:11Biomedical Informatics and Chair
• 00:13of the new Section of Biomedical
• 00:14Informatics and Data Science.
• 00:16Lucilla came to us from the
• 00:19University of California in San Diego
• 00:21where she was the Associate Dean and Chair
• 00:23of Bio, Chair of Department of
• 00:26Biomedical Informatics and she
• 00:28is going to describe some of the
• 00:30initiatives that BIDS is running towards
• 00:33broadening biomedical and clinical
• 00:36data access and utilization at Yale.
• 00:40Thank you, Leslie.
• 00:42Thank you. Let me share the screen
• 00:47and make sure it's the right one.
• 00:49We can see it looks great excellent.
• 00:52So thank you and and because it's a small
• 00:54group please feel free to interrupt.
• 00:56I might not always see your hand up
• 00:59but I asked the the organizers to help
• 01:02me if there is any question in the
• 01:05middle in and today what I will present
• 01:08is mostly what's in it for you at the
• 01:12Biomedical Informatics and Data Science.
• 01:13I'm not necessarily presenting
• 01:16our research or featuring a lot of
• 01:20projects that are ongoing right now,
• 01:23but it's essentially how you can
• 01:25take advantage of new things that are
• 01:28coming to enhance your own portfolio.
• 01:33So as you know informatics has been
• 01:35at you for a very long time.
• 01:38Since
• 01:4319, 1985 or 86 there has
• 01:45been a training grant.
• 01:47So you can imagine before that there
• 01:50was already activity in this area
• 01:52with the Yale Center for Medical
• 01:55Informatics as as well as many
• 01:58activities in Computational biology.
• 01:59But the faculty had appointments
• 02:02in different departments and and
• 02:04Yale was was not perceived as a
• 02:10large informatics academic unit.
• 02:14So we came to fulfill that that
• 02:18need and created the section,
• 02:21the first standing section of
• 02:23Biomedical Informatics and Data
• 02:25Science starting January 1st of 2023.
• 02:28So we're here for a little over
• 02:31a year and with the the mission
• 02:34of research training and services
• 02:36with the research with the faculty
• 02:38who are currently in department
• 02:41focusing on AI and medicine in
• 02:44our natural language processing
• 02:46including large language models,
• 02:48privacy enhancing technology,
• 02:51blockchain technology.
• 02:52In some examples of large projects
• 02:55that did transfer to Yale,
• 02:58the Centre of Excellence in
• 03:00Genome Science focused on admix
• 03:03genomes and also recruitment.
• 03:05A portion of the all of Us precision
• 03:09medicine program is is not here at
• 03:12Yale and we're in Puerto Rico as
• 03:15an associated entity in training.
• 03:19We have curriculum for Master's PhD students.
• 03:24We, we do mentor postdocs,
• 03:26clinicians moving into this area,
• 03:29biomedical data scientists and we'll
• 03:32have summer interns from North America
• 03:35and also from Africa starting this year.
• 03:38The new certificate program in medical
• 03:41software and medical AI has been advertised.
• 03:45The applicants have submitted their forms
• 03:49and we are going to start very soon.
• 03:54In terms of infrastructure and services,
• 03:56we are mounting the research,
• 03:58computing and clinical and biomedical
• 04:00data services that may be of interest
• 04:03to you and that's what I'll focus
• 04:06my presentation on and of course
• 04:09collaborating and grant proposals,
• 04:11quality safety research.
• 04:14We were fortunate to to get about
• 04:17$17 million this year of of grants.
• 04:21Since the multiple year grants we hope to
• 04:25have increased that in subsequent years.
• 04:28There were 19 grants in contracts
• 04:31and 27 new proposals by our faculty.
• 04:34So I'll talk mostly about the service
• 04:38portion which is staffed by our faculty.
• 04:41It's the Yale Biomedical Informatics
• 04:44and Computing
• 04:48Center, or CORE, in which we have
• 04:51a Research Information officer,
• 04:53and that's the Doctor Daniella Meeker in a
• 04:59Computational Infrastructure Officer with
• 05:01Assistant Dean for Informatics washroom.
• 05:04And the principle here is our section,
• 05:08it's one of several units that
• 05:11benefits from informatics,
• 05:12informatics, infrastructure,
• 05:13the Department of Medicine is another one,
• 05:16surgery, various departments,
• 05:21YCCI, the YCC and so on.
• 05:22And then we operate with partner entities
• 05:25such as the EU New Haven Health System,
• 05:28the EU University and so on.
• 05:31And we have to do that with compatible
• 05:36policies in essentially we're here
• 05:40to support research operations
• 05:42related to data and compute that
• 05:46tied together this this 3 units.
• 05:51So in terms of the department centres,
• 05:54every department will have
• 05:56an informatics liaison.
• 05:58Who is the main point of contact?
• 06:01Who gather helps prioritize needs
• 06:03of the departments and bring to us
• 06:06identify data sources that are needed,
• 06:09refer trainees,
• 06:10and partner in grant proposals
• 06:14or other activities.
• 06:16So let me get back to here
• 06:19that you see an initial set of
• 06:23Advisory Board members who help
• 06:26us prioritize what we need to do.
• 06:29Not all departments yet represented,
• 06:31but we we think we have a large
• 06:35coverage here and plan to expand.
• 06:38Now we ran an AI survey you you may
• 06:42have gotten this spam in our inboxes
• 06:45more than once and we wanted to survey
• 06:48what are the capabilities and interest
• 06:51in AI at the School of Medicine.
• 06:54So we had 1100 respondents from
• 06:58all categories and there's high
• 07:00interest in AI and in AI training.
• 07:04So on the right hand side you will
• 07:06see the number of respondents and
• 07:08then the on the line going across
• 07:11it is the percentage of them
• 07:14who were interested in training.
• 07:17So you can see it's it's all more
• 07:19than half of the respondents
• 07:21and in general great interest in
• 07:25ChatGPT in machine learning and
• 07:28language models and so on.
• 07:31Now we ran in two months we organized
• 07:36an AI in medicine symposium,
• 07:38the first one here at the School
• 07:41of Medicine in which we wanted also
• 07:44to showcase and actually honestly
• 07:47learn about what kinds of AI are
• 07:49being done in various departments.
• 07:53We one of the presentations was about
• 07:56that this model we we call Meiyama
• 07:59a medical large language models.
• 08:01LAMA is a foundation model produced by Meta,
• 08:07formerly Facebook,
• 08:08in which you can add things on top of it,
• 08:10do more training on top of an existing
• 08:14foundation model and then create
• 08:16medical large language models and
• 08:18that that was done by researchers from
• 08:22our team and there are others as well.
• 08:25So this is one of them.
• 08:27Another one is called Meditrom,
• 08:29also a faculty member in BIDS.
• 08:32In our unit there's ophthalmology
• 08:34GPT and this is done by a faculty
• 08:38who has joint appointments with us
• 08:40and with a secondary appointment
• 08:43in ophthalmology.
• 08:44And also there is GUT GPT and another
• 08:49large language models and and this one
• 08:53coordinated by by Dennis Chung and NGI.
• 08:57There are opportunities.
• 08:58So Doctor Kara has mounted this
• 09:02slide in in terms of engagement
• 09:05with healthcare that there are
• 09:07parts of it that we are currently
• 09:10not doing as adequately to get
• 09:12to the endpoint of having people
• 09:14interact with the health system.
• 09:17So he he mentions experts cannot
• 09:20consistently diagnose structural heart
• 09:22disorders and it's a hidden disease
• 09:25and we often only diagnose when
• 09:27it's at a more advanced stage.
• 09:30So there are apps and there are
• 09:34other AI based techniques that
• 09:36help diagnose earlier before the
• 09:40advanced treatment and evaluation.
• 09:43So here is a slide from his lab
• 09:47in which he puts on the left hand
• 09:50side the AICG applications that
• 09:52he has already put it out there.
• 09:55There's a wearable cards DS plus
• 10:00application as well and trials
• 10:02that have been done on top
• 10:04of machine learning models,
• 10:06again developed here at Yale.
• 10:12Doctor Sheng in GI is also several
• 10:16applications, one of them related to
• 10:20human algorithmic interaction with
• 10:23several collaborators and that uses that
• 10:26gut GPT application that I I mentioned.
• 10:29So a lot of interesting
• 10:32applications here at at Yale.
• 10:38Those two by the way are secondarily
• 10:41appointment with us and they're
• 10:43primarily appointed in the
• 10:45Department of Medicine at sections
• 10:47of cardiology and GI respectively.
• 10:49The BIDS faculty we have now 12
• 10:53ladder rank primary faculty and 13
• 10:56non ladder rank faculty and then
• 10:59again 40 people who were appointed
• 11:02in other departments and have
• 11:05secondary with us and those are in 20
• 11:09departments in five different schools.
• 11:11So what we see is on the on the left hand
• 11:16side BIDS having data workflow management,
• 11:19AI and methods expertise in
• 11:20the departments and centre 7.
• 11:22On the right hand side the data
• 11:24and the domain expertise for the
• 11:27particular application that we are
• 11:29talking about be that AI or other
• 11:31forms of informatics and data science.
• 11:34So in terms of modalities,
• 11:36we do have bioinformatics,
• 11:39imaging and clinical informatics and
• 11:42some are more populated than others at
• 11:45present and our plan is to keep moving
• 11:49and having this applications done on on.
• 11:53Again on the department and centre
• 11:55side we find many people who
• 11:58have been doing informatics on a
• 12:00fragmented way and would try to not
• 12:03only acknowledge what they're doing,
• 12:05but try to synergize and create
• 12:09the computational infrastructure
• 12:11that it can again can get to high
• 12:15higher levels because there are more
• 12:18people using them at this point.
• 12:21So our goal of developing new computer
• 12:24based methods and systems to improve
• 12:26human health is part of our mission to
• 12:30train the next generation and then help
• 12:32researchers make maximal use of data.
• 12:35So I would expect many today in
• 12:38the audience are working with the
• 12:40data from the health system or from
• 12:43clinical trials or from basic science
• 12:46and in that all data management
• 12:49in advanced methods may benefit.
• 12:54So AI is the current interest of
• 12:58of many people and there are NIH
• 13:01requests for applications in many,
• 13:03many different areas in institute
• 13:06that puts them here as examples.
• 13:08And then on the left hand side,
• 13:10I select some topics that are very,
• 13:14very related to it as well as projects
• 13:17that we are currently involved.
• 13:20One is aim ahead the HealthEquity and
• 13:23advanced diversity for the workforce.
• 13:26The other one is Bridge two AI in
• 13:29other common fund programs and and
• 13:32as I mentioned before the all of
• 13:35Us Precision Medicine initiative
• 13:37AI is a team sport.
• 13:38So again I put both both sides here so
• 13:42that we don't exist on methods only
• 13:45we have to have you know important
• 13:48applications and have the right
• 13:50questions to answer and that's where
• 13:53regardless of the medical specialty
• 13:55we we do need partnerships and help.
• 13:59So data,
• 14:00computational infrastructure and training
• 14:02is what we're we're trying to do.
• 14:05And and here's a busy diagram in
• 14:09terms of the computing infrastructure
• 14:11that we found
• 14:13needs to be developed at Yale so
• 14:16that we can have everyone fulfill
• 14:18their application needs as well
• 14:21as new development of methods.
• 14:25So going from from #1 to #4,
• 14:28what safe is a computational
• 14:31infrastructure designed to have HIPAA
• 14:34compliance and to have protect sensitive
• 14:38information such as electronic health
• 14:41records and clinical trial data and
• 14:44that has a certain amount of storage
• 14:47that that needs to increase from.
• 14:50What is that?
• 14:51There is already a system in there,
• 14:53but it it needs a lot of updates for
• 14:58that and the research Virtual Display
• 15:02Interface is how you use your own
• 15:07computer to serve as a terminal to
• 15:10servers that are much more powerful
• 15:12and again secured for this operation.
• 15:14So you you use your computer to access data
• 15:19and to do analysis on this protected server.
• 15:23So, so this again is is being enhanced from
• 15:27an existing seed compute infrastructure
• 15:31that existed and then you have the cloud,
• 15:35the Amazon Web Services spin up plus
• 15:39which will be more secure version of
• 15:43the current capability that exists
• 15:46throughout the university.
• 15:48This will be possibly adequate
• 15:51for genome annotation,
• 15:53genome analysis and image processing as well.
• 15:58And then finally on the right hand side,
• 16:01it's what we are calling Microscope
• 16:03High Performance Computing.
• 16:05And this is a set of servers with GPU nodes
• 16:09and these are graphic processing units
• 16:13where they're necessary to do AI these days.
• 16:17And it is something that currently is
• 16:20not available to to our community.
• 16:25So why do we do all that?
• 16:26Well,
• 16:27because we we want to use data.
• 16:28And in the past I've
• 16:30created data networks from
• 16:34several institutions,
• 16:35including the national VA,
• 16:38so that we could use more data because
• 16:41however large our system can be,
• 16:45there are important other diversity
• 16:47of data in in other areas.
• 16:49So that's something we want to
• 16:52eventually participate in this clinical
• 16:55data networks with participation of
• 16:58the Yale electronic health records,
• 17:01clinical data and so on.
• 17:02We're we're not there yet.
• 17:04There's a lot to be done in data
• 17:06curation and preparation but but we're
• 17:09working towards that goal also in
• 17:11terms of precision medicine and and
• 17:14PRS which has polygenic risk scores.
• 17:17We we do think we need to pay attention
• 17:21to that particularly because the
• 17:24current studies are lacking diversity
• 17:28in the population and they're
• 17:31incorrect findings being reduced.
• 17:33So can we play a role in that
• 17:37with again the diversity of the
• 17:40local population participating
• 17:41in in more studies and and then
• 17:44doing this large networks as well?
• 17:46Can we do that without introducing
• 17:49more biases and can we protect privacy?
• 17:52How do we study individuals from mixed
• 17:56ancestries who are currently many
• 17:58times discarded from analysis because
• 18:02it's harder to do the competition with them?
• 18:05So we have a A Center of Excellence
• 18:08in Genome Science based on Admixture
• 18:10Center of Admixtured Science and
• 18:13Technology in which we want to
• 18:16account for particular ancestry
• 18:19based on the specific chromosome and
• 18:23location rather than global ancestry
• 18:26as it's currently done in genetic
• 18:29studies because we know there's
• 18:32improve power if we include that
• 18:35mixed populations in the studies.
• 18:39So just schematically A trait is
• 18:43influenced by genetic determinants,
• 18:46exposures,
• 18:46social determinants and within genetic
• 18:49determinants there are variant effects.
• 18:53There is also ancestry effects and
• 18:57we need to account for for all of
• 18:59that and the way we do things,
• 19:01unfortunately or fortunately,
• 19:03that the data are sequestered
• 19:06in different compute enclaves.
• 19:09Fortunately because that helps protect
• 19:12the security and privacy of the data,
• 19:16but unfortunately because it makes
• 19:18the calculations way more difficult
• 19:21to to be done.
• 19:22But we have technology workarounds
• 19:26that allow us to compute with both sets
• 19:29of data that this will be eventually
• 19:332 million whole genome sequences
• 19:35that we can use to do precision medicine.
• 19:39So we have work in this area of
• 19:42secure Federated algorithms in
• 19:45which currently we are working
• 19:48with computing with all of U.S.
• 19:50data and the million veteran program data.
• 19:53But we believe that as new Biobanks and
• 19:58new programs appear in other countries,
• 20:02if they follow a similar protocol
• 20:05that we could eventually do
• 20:07that with other nations as well.
• 20:10And that would greatly increase
• 20:12the power of all the analysis.
• 20:15And again,
• 20:16these are algorithms that allow
• 20:19all participant data to stay
• 20:21in their compute enclaves.
• 20:24Kindness of questions that the
• 20:26precision medicine programs such as
• 20:29all of us try to answer once again,
• 20:32all all sorts of them and and several
• 20:35different specialties, right so.
• 20:37So this was a slide from the
• 20:39program that helps to disseminate
• 20:42and try to convey the importance of
• 20:45doing a cohort like this where we
• 20:50have electronic health record data,
• 20:52survey data, physical measurement data,
• 20:55bio samples in a a small part,
• 20:58wearables and digital app data as well.
• 21:02So we are recruiting here
• 21:05in New Haven. We have at
• 21:11YCCRAYCCI a recruitment portion
• 21:14and also in in a few other places
• 21:17in we partner with the Precision,
• 21:22the Puerto Rico Center for
• 21:24Clinical Investigation as well.
• 21:26We have a recruitment site in there,
• 21:29so adding to to the program.
• 21:32And then another thing that I I
• 21:35wanted to talk about is because it's
• 21:38being launched as as we speak which
• 21:41is the Cosmos community data set.
• 21:44Cosmos is an initiative from Epic of the
• 21:47vendor of electronic health records system.
• 21:50In that if a particular customer of
• 21:54EPIC such as us wants to participate
• 21:58then you would have access for de
• 22:02identified data for research not
• 22:05only of our own data here at the EU
• 22:09New Haven health system but also of
• 22:12other institutions participating.
• 22:14So right now there are 230 million
• 22:18patients represented with several
• 22:20encounters and face to face visits.
• 22:23So it's it's about 1300 hospitals
• 22:27so quite a a a very large group
• 22:31of them and and you can use this
• 22:34data for research
• 22:38the analysis tools the one that
• 22:41is available for everyone is
• 22:43Slicer Dicer on this larger set.
• 22:46Of course you you already know of
• 22:49Slicer Dicer for the electronic
• 22:51health record systems for the epic
• 22:54system here at the health system.
• 22:56But this would be you can as as
• 23:01we get to this launched use Slicer
• 23:04Dicer as A to do research or to do
• 23:09some aggregate in graphics on the
• 23:13full population which is about 200.
• 23:16It's above 200 such health systems
• 23:20represented so much larger than than our own.
• 23:24And then the analysis tools.
• 23:27You can also do command line R,
• 23:30Python, SQL statements,
• 23:33but that requires certification that it
• 23:36requires that you you were trained in
• 23:39in the particular aspects of doing that.
• 23:42So what we're doing right now is
• 23:46to start with a class of 1 or more
• 23:49representatives for different departments
• 23:51so that they can be trained on this,
• 23:54and then start helping fulfill requests
• 23:58that come from their own departments.
• 24:02Their data management tools are listed here.
• 24:05And then code libraries as well.
• 24:07And some of you may recognize some of
• 24:11these libraries used for AI type research.
• 24:14The good thing too is that this whole
• 24:18computer environment is already
• 24:21existing is from the epic side.
• 24:24So while we build ours,
• 24:27there is one that can already be
• 24:29used by data analysts.
• 24:34Now the the timeline is like I said,
• 24:36we are right around the,
• 24:38the corner of being available
• 24:42in terms of Cosmos Live meaning
• 24:45this slicer Dicer portion of it.
• 24:48And then we we have people doing
• 24:52the prerequisite completion for the
• 24:54course that they will take April 8:00
• 24:56and 9:00 in order to be certified
• 24:59to be used for that platform.
• 25:02There will be 30 initial people
• 25:05trained and and then monthly
• 25:07we will add a few more.
• 25:12So in summary, what we're doing for the
• 25:15informatics infrastructure for research
• 25:17that you all benefit from is catching
• 25:21up on hardware and cloud security,
• 25:25cloud environments and also on the data,
• 25:28on preparing the data for
• 25:31analysis we are are in,
• 25:34we have some initial work and policies
• 25:36and and the training of people,
• 25:38we have much more to do in this area,
• 25:41development of software and then the
• 25:43launching of services in in the overall
• 25:46training is also in the pipeline.
• 25:48But but we decided to go one thing at a
• 25:54time because we were still a small group
• 25:56as I mentioned before the researchers
• 26:00in the case we're recruiting faculty
• 26:02not only to do their own research
• 26:05and be users and testers of this
• 26:07environment but also those who lead
• 26:09the development of this whole services.
• 26:12So, so that has been an
• 26:14area of emphasis of ours,
• 26:17research scientists and postdocs.
• 26:18We need to have PhD students,
• 26:21medical students, undergrads,
• 26:22the whole community of researchers
• 26:25to to benefit from this environment.
• 26:28But the environment is still coming up,
• 26:30which is why I wanted to present here
• 26:34and then also answer any questions
• 26:38related to requirements or you know,
• 26:43interest in external data sets as well.
• 26:46Not just the electronic health
• 26:48records from EPIC,
• 26:49but some researchers work with claims
• 26:52data that are licensed from other places
• 26:57and some many other sources exist.
• 27:01So.
• 27:01So with that I'll stop sharing and see
• 27:04whether there are any questions or comments,
• 27:08suggestions on on how what you need
• 27:10to do most so we can move ahead.
• 27:22So I think one question I had was
• 27:23you know if they're faculty that are
• 27:25interested in working with bids is there,
• 27:27how can they reach out?
• 27:29I mean should they reach out to you,
• 27:30is there an online form or how
• 27:32does that work it it depends.
• 27:36Typically if they aren't like in those
• 27:39informatics people who happen to just
• 27:41be in another department but want to
• 27:44affiliate with us then they they talk
• 27:46to me we we set up a a time go over
• 27:51as secondary faculty what they they
• 27:53would like to benefit from and and so on.
• 27:56So those are the informatics
• 27:59faculty appointed elsewhere,
• 28:00the informatics users who
• 28:03want for example data.
• 28:05The whole JDAT.
• 28:08Currently,
• 28:08JDAT Group is being revamped and
• 28:13renamed and reorganized so that the the
• 28:18data requests can be more streamlined
• 28:20than they are right now and actually
• 28:23filtered because some of them are not,
• 28:28you know,
• 28:28at at the same level as as others.
• 28:30In terms of how researchers have thought
• 28:35about the request and whether electronic
• 28:39health records can really answer the
• 28:41question that they have in mind.
• 28:43Because many times it's not.
• 28:45The question you have is the
• 28:48question that's electronic
• 28:50health records can help answer.
• 28:52So there there is a lot of re
• 28:55restructure in that area in terms
• 28:59of compute against some faculty.
• 29:02But those are typically more the the
• 29:05the informatics ones have computer
• 29:07needs that we currently cannot fulfill.
• 29:10But we're designing the new structure
• 29:14in order to fulfill anything also
• 29:19related to security of people who
• 29:23might somehow have inherited data
• 29:25sets in local servers or laptops or
• 29:29other things that need to be moved.
• 29:32That's the whole area in which the
• 29:35Chief Research Information Officer,
• 29:37Doctor Meeker is is working towards,
• 29:40because we do have to move certain data
• 29:44sets into environments that are more
• 29:49up to date with regards to to security.
• 29:57Yeah, I have. I see a doctor
• 30:01chalk. Thank you, Doctor
• 30:04Ono Machado for for a nice talk.
• 30:07I just have one very short question.
• 30:10You mentioned about the All of Us
• 30:12precision medicine research program.
• 30:14Is that immediately available to work on?
• 30:19Can you speak a little
• 30:20bit about that? Thanks.
• 30:22Yeah. So the data for the All
• 30:24of Us program is immediately
• 30:26available and in fact several Yale
• 30:28researchers have already published
• 30:31based on on data from that program.
• 30:33And the reason it's very timely
• 30:36right now is one they also provide
• 30:39their own computer environment.
• 30:41So we don't need to provide ours.
• 30:45You you essentially go online and
• 30:49as long as you have an area Commons
• 30:52account you can sign up for it.
• 30:55There are a few
• 30:58online training that you
• 31:00have to to go through.
• 31:01So there is some prerequisites in
• 31:04order to access the environment
• 31:06and then and that's for the
• 31:09environment similar to Cosmos,
• 31:10the environment that you do command
• 31:12line that you select cohorts and so on.
• 31:14If you just want to do a data
• 31:17browser so similar to Sliced Redicer,
• 31:19you can do it immediately
• 31:22without registration.
• 31:23The very interesting portion of
• 31:26it is the genome sequencing,
• 31:29which right now is already 350,000
• 31:34and that's already one of the
• 31:36largest collections smaller
• 31:38than UK Biobank currently.
• 31:41But in terms of diversity,
• 31:43is is not even comparable because
• 31:46the whole recruitment process,
• 31:49the whole program is predicated on
• 31:52eliminating that bias that exists
• 31:55that genetic studies are based on
• 31:59European populations for the most part.
• 32:08Yes, Doctor,
• 32:10wonderful, wonderful summary.
• 32:11Thank you so much for sharing.
• 32:14So I'm Kim Blendman.
• 32:16I am an assistant professor here
• 32:18in Medical Oncology at the School
• 32:20of Medicine as well as an assistant
• 32:22professor in Computer Science at
• 32:23the School of Engineering and
• 32:24Applied Science here as well.
• 32:26And I was very much so interested
• 32:28in your Cosmos community data set.
• 32:30And you mentioned that you know you
• 32:33will allow some of your faculty as
• 32:35a assistant myself to be, you know,
• 32:37users and beta testers for some of the
• 32:39things that you have moving forward.
• 32:40And I was just wondering, you know,
• 32:42how do we follow up on that to try to
• 32:44get a little bit more information of,
• 32:45you know, what's, you know,
• 32:47more deep dives into what's actually
• 32:48in the data set and things like that,
• 32:50that we could, you know,
• 32:51as faculty, you know,
• 32:52jump into and and and and query and you know,
• 32:56understand a bit more of how we
• 32:58can use it in our for our research,
• 33:00our personal research,
• 33:01Right. Because the seats were
• 33:04limited for the class of 30,
• 33:06we had asked the department chairs
• 33:09to nominate who they wanted.
• 33:13And we actually did not encourage
• 33:16faculty to be the the people necessarily
• 33:19because whoever is first has to serve
• 33:22the needs of a whole lot of others.
• 33:25So we even recommended being
• 33:27staff members who were highly
• 33:29skilled in the data analysis,
• 33:31but in some cases it was, you know,
• 33:34there wasn't anyone with that description
• 33:37and they wanted faculty and and so we
• 33:41we fulfilled one per per department.
• 33:45So. So the moving forward there will
• 33:48be at at least two to four each month
• 33:51who will undergo the training as well.
• 33:54So I mean the goal is one day
• 33:58everyone who has the the skills
• 34:00to to do SQL queries plus R or or
• 34:05Python as a programming language
• 34:09would be would have access.
• 34:12It just takes some time because the
• 34:15epic does require the training and in
• 34:18fact I like that because it it becomes
• 34:20not a completely open environment
• 34:23for everyone who would because we
• 34:26all know that even de identified
• 34:29data can be easily DE identified.
• 34:32So there is this filtering of of
• 34:37people's not everyone in it also
• 34:39it's only individuals from the
• 34:42institutions that are contributing data.
• 34:44So there is control over,
• 34:46they are employed in this institution.
• 34:48If something is done wrong,
• 34:51wrong way to try to re identify and so on
• 34:55then there could be consequences and so on.
• 34:58So I think it's the,
• 35:00it's the first step towards I'll say
• 35:03opening the data a little bit and if
• 35:07that works then there will be other
• 35:11offerings at the same time as
• 35:14we develop the computational
• 35:15infrastructure that we need. No,
• 35:18I I totally agree with that.
• 35:19You know how you're you're opening
• 35:20this up in terms of you know
• 35:23individuals who would be really to
• 35:25serve as that that that core role you
• 35:27know where they have the bandwidth.
• 35:28And yeah I agree that it wouldn't be
• 35:30you know as appropriate to have the
• 35:31faculty do that but are there you
• 35:33know so you mentioned that you're
• 35:34the J dad has is being restructured
• 35:37and that you know could you and you
• 35:39spoke a little bit about you know
• 35:41you know the offerings you know that
• 35:42it will have as is moving forward.
• 35:44Do you have a thought in terms of when
• 35:46that will be rolled out and when we can,
• 35:48you know, start to connect with J Dad
• 35:51again about, you know, pulling out,
• 35:53you know, information and things
• 35:54of that that nature that we can,
• 35:56you know, get get things moving again?
• 35:59Yeah, it hasn't stopped this way.
• 36:03But what happens is it's, you know,
• 36:07because the volume is so high,
• 36:10too long and maybe even when it
• 36:12comes then the people don't need
• 36:15the data anymore so that it causes
• 36:18even more waste of resources.
• 36:20So what we're trying to do is to the,
• 36:24the one upfront moving forward when
• 36:27when we launched this, this new era,
• 36:30right, do a feasibility analysis and
• 36:33and are this data really you know,
• 36:37can we even solve that problem?
• 36:39There are several trainee questions right?
• 36:42Medical students, postdocs,
• 36:44graduate students who and currently
• 36:48there there isn't a prioritization.
• 36:51It's you know everyone enters the
• 36:54queue and I think that's another
• 36:57area in which we will have to think
• 37:01about more carefully in in other
• 37:05institutions the cost self selects who
• 37:09who is putting those requests or not.
• 37:13Of course we don't want that.
• 37:15We don't want to prevent good
• 37:18ideas to to move forward.
• 37:20But we can also not have any
• 37:24mechanism to to determine that.
• 37:27This particular research question is not
• 37:32well answered with one single system,
• 37:37you know observational or
• 37:40retrospective analysis.
• 37:42It's it just would require much more
• 37:45which then again having opportunity
• 37:47to use all of us is an opportunity
• 37:51to use Epic Cosmos as well.
• 37:55So.
• 37:55So we want to move towards a point
• 37:59that we can decentralize somewhat
• 38:02this request because at least
• 38:05the feasible feasibility we no,
• 38:07we don't have enough patience with this,
• 38:10this and this.
• 38:12We're not going for that trial
• 38:15because it's not feasible.
• 38:18But that all requires a lot of work
• 38:22because since the data haven't been
• 38:25used as much and you only encounter
• 38:28data issues as you use the data, right?
• 38:33Absolutely.
• 38:34There will be a lot that we'll
• 38:36have to do as we start opening
• 38:39this new possibilities.
• 38:43Now, what was the what was
• 38:44the new name of the J dot?
• 38:45You said that it's changed names as well.
• 38:48Oh, it's still under the bait. OK All
• 38:53right. Thank you so much.
• 38:54This is fantastic. I, I really,
• 38:56you know, love what you're,
• 38:57what you're doing and how you're
• 38:59reorganizing it, you know,
• 39:00to make it more streamlined
• 39:02and more accessible to people
• 39:03who who don't have that,
• 39:05that background and skill.
• 39:06And I hope that those of us
• 39:07such as myself who do have the
• 39:09background and skill can, you know,
• 39:10be a part of this movement as well.
• 39:12Oh, exactly. And I I want to thank
• 39:14everyone for the patience because yes,
• 39:16we've been here a year in two months now and
• 39:23it is not as simple,
• 39:29yeah, but but it's doable.
• 39:31Others done it.
• 39:32We can do it to edit it at UCSD many,
• 39:36many years ago. And so again,
• 39:41if we can navigate the the aspect
• 39:46of the cultural aspect on on trust
• 39:50and and how we we move this along,
• 39:53I think we'll go, we go a long
• 39:56ways just taking a little long. Do
• 40:18you have any more questions?
• 40:26No, Thank you so much for coming.
• 40:28Oh, Eugenia, do you have another question?
• 40:31No, no. Thank you so much for
• 40:34coming and presenting about bids.
• 40:35I think everybody really enjoyed it and
• 40:36it's exciting to see what there is to come.
• 40:40Thank you so much. Bye, bye.