Dear "member" of the OpenTox community:

With great pleasure I am looking back at our OpenTox Euro meeting in 
Dublin. I really think it was a great meeting with many excellent talks 
and productive discussions and I hope that you now see yourself as part 
of OpenTox. My special thanks goes especially to the group participating 
in the working group discussions on the first day. Even if we had some 
virtual meetings before, I see this date as the real staring point of 
the working group activities. What was planned as four separate meetings 
ended up to be one meeting of the complete group since we directly 
recognized that the topics of the four working groups are extremely 
interlinked. There is need for strong interactions with the AOP 
activities leaded by Clemens Wittwehr (WG2) in two directions: 1) Data 
from the OpenTox community should be made available to verify APOs and 
2) information from AOPs should be included in OpenTox modelling tools. 
This is only possible with standardized data/metadata (WG3, Thomas 
Exner) and extended APIs (WG1, Christoph Helma). All this has then to be 
provided in a user-friendly environment (WG4, Tim Dudgeon). The 
following discussion showed that there are many different opinions and 
demands on interoperability and standards. To get all these into one 
working model is definitely not an easy task. To approach it, two main 
goals for the near future were proposed:

1) Documenting the past: Unfortunately, the openTox developments are 
very fragmented at the moment. To change that, Tim volunteered to 
generate an overview of existing OpenTox services. You can find more 
details in his E-mail below. Besides the services, which resulted from 
the OpenTox projects and its successors, we are also welcoming all 
suggestions for services, which should be linked to OpenTox. Especially, 
the input from Yanli Wang (PubChem) was extremely helpful in this 
respect and we are looking forward to a fruitful cooperation.

2) Planning for the future: Even more important than file format 
standards for the data and metadata are quality standards of the data. 
Only if we can judge the quality of the data, we can decide if we want 
to use them in modelling or if it is acceptable for regulatory purposes. 
But already these two applications demand for different information 
stored as metadata. Is it possible to design a infrastructure, which 
stores all data/metadata, so that it can easily be retrieved and 
included in decision making or converted to reports like REACH and SEND? 
To find this out, use cases or better backbones of uses cases will be 
developed and analyses regarding their demands on the data formats. In 
Dublin some more volunteers were identified:
a) Clements: Data for regulatory purposes
b) Christoph, Thomas: Datasets for tox prediction
c) Thomas, Joh Dokler (Douglas Connect) (and Yanli Wang): Data exchange 
using PubChem as most important example
d) Roland Grafström, Hristo Alajdov, Clements, Thomas: Interacting and 
modelwith AOPs
e) Ignacio Gonzalez Suarez (PMI): standards for high contents screening
Maurice Whelan (EC) pointed us to the successful use of use cases in the 
SEURAT-1 program including ToxBank and ask us to take a look at them, 
which we definitely will. To cover even more opinions, IO would like t 
ask you to propose additional use cases. These don't have to be fancy 
but just describe your daily life --> How do you use tox related data 
and what information you need or would like to see associated with this 
data. Just a short draft of the work flow is enough highlighting the 
problems in current datasets. I will create a doodle (additional 
invitation will follow soon) to find a date for a follow-up virtual 
meeting end of this month in which the use cases can be discussed.

Thanks again for all the valuable input on the working groups.
Thomas


______________________________________________________________________
Tim Dudgeon:

Hi All,

Here are some thoughts on what we can do to create a useful directory of 
OpenTox services.
I'd welcome feedback on this.

1. List of services
See first sheet here:
https://docs.google.com/spreadsheets/d/13RlxLQWunwJAvvjPQWDSciokUnZQhz0vCQyMyBt4jNI/edit#gid=0
Whilst a useful start, its probably not that useful at present.
It would benefit from a more complete description of each service, 
including a description of the methodology behind the prediction 
services, and other aspects that differentiate the services.
And maybe some other columns could be useful (please suggest).

2. List of models
A service on its own isn't much use to a user. Its presumably the 
ability to use a model within a service to generate predictions that's 
of real use. Therefore it would be useful to somehow catalog the 
available models in each service. But new models can be created on the 
fly, and just because a user has created a model doesn't mean its any 
use, so maintaining a useful catalog for this across all services would 
be almost impossible.
But maybe we can handle this if we treat this on the basis of "approved" 
models, which have been generated from "approved" datasets (see next 
section) and have been "validated" (in the sense of having been shown to 
generate reasonable results).
Do you think this is possible/reasonable to try to achieve?

3. List of datasets
If predictions are to be comparable between different services then they 
need to be based on the same datasets.
Therefore it might be useful to prepare a list of "approved" datasets 
that can be use to make models, and to have those datasets available in 
standard formats.
This way each service could generate predictive models using these 
"standard" datasets, and this can allow different models to be compared. 
These could form the core of the "List of models" described above.
These standard datasets would not be restricted to those within OpenTox, 
but would include ones from ToxCast, Tox21 and other places.
Do you think this is possible/reasonable to try to achieve?


I welcome any thoughts on any of this.

Tim



_______________________________________________
Alllists mailing list
Alllists at opentox.org
http://www.opentox.org/mailman/listinfo/alllists