OHDSI Home | Forums | Wiki | Github

Access to the data network


(Ioan Filip) #1

Hi! I am postdoc and I just joined the OHDSI forums. I would like to explore OHDSI’s data network in order to familiarize myself with the CDM for research. I am interested in applying new methods in Topological Data Analysis to this data bank. How can I obtain access to the network?


(Christian Reich) #2

@Ioan_Filip

Welcome to the family.

You have to ask the data holders. It’s decentralized. We don’t have one Czar who can give you access to all sorts of data.

What kind of data are you interested in and what kind of topology do you want to study? Maybe we can help you point to somebody with the right stuff.


(Ioan Filip) #3

Hi Christian! Thanks for the clarification.

I am interested in the following broad topics:

  1. infectious disease modeling;
  2. the transcriptomics of viral infections, as well as the host responses;
  3. any biological correlates from medical records and patient surveys
    related to the immune system;
  4. whole exome and genome sequencing, and their associated phenotypic
    correlates.

Our lab has a focus on cancer research, GBM in particular, and cancer data
sets are also of great interest.

The topological methods that may be of use are the Mapper algorithm, which
provides a dimension reduction and clustering procedure that is more
sensitive to the metric structure of a point cloud than typical methods,
and persistence homology, which can extract meaningful features related to
the connectivity of a point cloud.

I would greatly appreciate your suggestions and recommendations about the
particular WGs and data sets within OHDSI!


(Christian Reich) #4

@Ioan_Filip:

OMG! How many PhD students really behind the “Ioan_Filip” moniker? Like 10? :slight_smile: This is the work of a whole department. You are truly ambitious. Well, good luck with your studies.

All of these use cases are very useful and interesting, and the OMOP/OHDSI system will allow you to study them. Well, the phenotype aspect, that is. We currently don’t have any real support for omics, even though that bubbles up as a future direction all the time, When you talk about point clouds and clustering procedures - are you talking about geographical distribution of effects? There are folks work on this already. You may want to hook up.


(Ioan Filip) #5

Thanks for the link.

I am using clustering in the general statistical sense, applicable not
necessarily for geographically located data points, but for any sample set
in a possibly large-dimensional space equipped with some notion of
similarity or distance.

There are several people in our lab interested collaborating with the
groups at OHDSI, so I mentioned all the areas that our group focuses on.

So if I understand correctly, all the data sets at OHDSI could be related
to these research areas I mention. And to access any particular one, I
first have to get permission from the corresponding WG which maintains the
particular data set.

Where can I find the contact information of the data holders for each
database listed on OHDSI?

Thanks in advance!


t