---
# UK TRE meeting - September 4th 2023

**Date**: 4th September 2023 10:00 - 17:30
**Zoom link**: https://swanseauniversity.zoom.us/j/91561196834?pwd=OGtVS1I4WU80ODk5MVY0U09TNVYvZz09
We are 200 people registered for today!
- 100 in-person
- 100 remote

## Useful links
### For today
* **Event info**: https://www.uktre.org/en/latest/events/wg_workshops/2023-09-04-september-meeting/index.html
* **Mentimeter**: https://www.menti.com/al2ycq6om535
* **Slido**: https://app.sli.do/event/jfazwcmCSwUcnWNpSthofE/live/questions
* :warning: HELP!: David at [e-mail] or David Sarmiento on RSE slack

### For the community
* **JISC Mailing list**: https://www.jiscmail.ac.uk/cgi-bin/wa-jisc.exe?SUBED1=RSE-TRE-COMM&A=1
* **Website**: https://www.uktre.org/en/latest/index.html
* **Slack channel**: https://ukrse.slack.com/archives/C045ETUPPD0
* **Report from last meeting**: https://www.uktre.org/en/latest/events/wg_workshops/2023-03-29-march-meeting/index.html
* **Newcastle Commitment**: https://rse-tre-community.readthedocs.io/en/latest/newcastle-commitment/index.html
* **Community GitHub**: https://github.com/uk-tre

## Schedule

| Time | Activity | Link to notes |
| ---- | -------- | ------------- |
| 10:00 - 10:15 | Arrival | N/A |
| 10:15 - 10:30 | Welcome and introductions | [Attendee registration](https://hackmd.io/ex6PZLPfQwat1vu1pDQ6cA) |
| 10:30 - 11:15 | Keynote from Emily Jefferson (CTO, HDR UK), "Can convening a technology ecosystem help TREs to work together?", including discussion | [Shared notes](https://hackmd.io/ciU0I3M6RqK95abJBJ5PwA) |
| **11:15 - 11:30** | **Break** | |
| 11:30 - 13:00 | Lightning talks | [Room 1 shared notes](https://hackmd.io/J6V-24saR9a8cZY1cTiuRA) |
| | | [Room 2 shared notes](https://hackmd.io/QTOlZnUQTeWVBLHCY0kt8w) |
| **13:00 - 14:00** | **Lunch** |[Online lunch chat](https://hackmd.io/MIKIdUkqRNCv56p8NxxzSA) |
| 14:00 - 17:15 | Parallel breakout sessions | [Template notes](https://hackmd.io/7Wb_wn7SQZmltfIe4VG5lg) |
| 17:15 - 17:30 | Closing remarks | [Shared notes](https://hackmd.io/JEdImRCrQ8is2yAKLEa-rQ) |

## Breakout rooms - possible topics
*Please add more topics below you would like to discuss in the afternoon, if they are not already here!

Link to vote: https://www.menti.com/al2ycq6om535

* [Governance of the UK TRE Community](https://hackmd.io/@manics/rknZ8Im0n/edit?both)
* [Multi-TRE analysis: challenges, governance requirements, federation](https://hackmd.io/@manics/HyA-LLQ0n/edit?both)
* [Current state of the art re data linkage/federation/AI&ML&LLM across infrastructures: federation, governance, safe output methods](https://hackmd.io/@manics/Sk1ML8mA3/edit?both)
*[Separate page for table 2](https://hackmd.io/y-CpfeT5TJ6jwlYrVv8j2g)
* [TRE sustainability and operations](https://hackmd.io/@manics/BklfLU7Rh/edit?both)
* [Safety and security of Python and R package import into TREs](https://hackmd.io/@manics/SkZGLL702/edit?both)
* [Community-based efforts and collaborations in public involvement and engagement](https://hackmd.io/@manics/SJzMLU7A2/edit?both)
* [Cloud vs on-prem TREs: costs, constraints, pros & cons](https://hackmd.io/@manics/rkEzLI7Rh/edit?both)
* [The role of independent TRE providers in relation to the NHS national and regional SDEs](https://hackmd.io/@manics/SkSzIUQRh/edit?both)
* [The role of TRE standards for the TRE community](https://hackmd.io/@manics/ByUfULXCn/edit?both)
* [Sight unseen: how far can we go with keeping data hidden from users?](https://hackmd.io/@manics/ByuMILXR3/edit?both)
* [Addressing data harmonisation between different datasets: do TREs have a role?](https://hackmd.io/@manics/HJFMLU703/edit?both)
* [Access and rights metadata to support process choreography and interoperability (e.g. generic researcher applications and authorization)](https://hackmd.io/@manics/rJ5MUUQAh/edit?both)
* [Future governance of the SATRE Specification](https://hackmd.io/@manics/Sksf8LQRn/edit?both)
* [Accrediting people as well as environments… will developing “trusted researchers” allow us to do better/more/quicker data science?](https://hackmd.io/@manics/BypML8QC2/edit?both)
* [What are the detractors to standardisation and working together? How can the reasons for the tension work alongside standardisation?](https://hackmd.io/@manics/Bk0GIImRh/edit?both)
* [There is a tension between vague scientific questions and the data reality. How do we better educate PIs in what TREs can and can't do for them and the journey ahead of them?](https://hackmd.io/@manics/B1ymUImCh/edit?both)
*
*
* [General discussion](https://hackmd.io/@manics/HkbznLQAh/edit?both)

Live results: https://www.mentimeter.com/app/presentation/almku17vy1fmvcpys3on9ju52nkoyit9

## Collaborative note
Please do contribute to the notes below.

## Welcome / intro

- Key questions
- What's stopping TREs from working together?
- What's actuallu happenign in the TRE space?
- What can TRE teams do together?
- What support can the UK TRE Community provide to help TREs work together?
- (What's missin?)

## Keynote
Can convening a technology ecosystem help TREs to work together? by Emily Jefferson (CTO, HDR UK)

- Goal to make it easier for accredited researchers to access data across TREs.
- Four pillars:
- Technology services ecosytem
- Trust and transparency
- Usable data
- Capacity building (RSEs a key aspect of this)
- Driver programmes focussed on answering research questions but with a brief to do so in a way that contributes to the collective pool of technology and approaches
- Common challenges / pipeline elements across TREs / projects, but little re-use / shared use.

- Convening a technology ecosystem
- Community, People, Beyond Health, Standards for Interoperability, Science of Infrastructure
- How can HDR UK support this?
- HDR's Technician Commitment - what's a better name?
- Recognising non-traditional careers including RSEs
- Supporting a wide range of impact - not just academic publications
- Technical solutions but for health need to work for non-health, too. People who work with sensitive data - health or not -> DARE UK
- Help us develop standards - OMOP, but what else?
- Innovative infrastructure which is making an important contribution - recognising impact.
- Technological Solutions
- Solutions from across the community
- HDR UK will also contribute some technological solutions supporting the ecosystem - all being open sourced. Not about mandating, but about having something available for all to use / build on.
- "The Gateway" to signpost and connect things across the communty and include some common core elements.
- e.g. Metadata dictionary, Cohort discovery, Data discovery, Data use register, Data use requests
- Mk2 Gateway: Brandable by others, interoperable, modular, community developed, re-use existing solutions / components, driven by use case exemplars, open APIs, automated, scalable

## Lightning Talks (Room 1)

- Implementing a TRE at Swansea - challenges and thoughts
- as a central IT service, separate from health centre
- stems from digital security policy and cyber security improvements
- built on AWS
- challenges
- small team of two people
- transition from on-prem to cloud
- bespoke computing resource
- capacity planning
- communication between stakeholders - needs to become more formal
- messaging and signaling
- How we manage LASER - Leeds Analytics Secure Environment for Research , the TRE at Leeds Institute for Data Analytics
- an Azure based TRE
- how do researchers get software into the TRE?
- Frog Artifactory
- Power BI dashboard for Project costs
- SATA Semi Automated Risk Assessment of Data Provenance and Clinical Free-Text in TRE.
- collab between several Scottish Safe Havens.
- risk assessment of clinical free-text by applying NLP to expore the data
- privacy risk dashboard for clinical free-text
- semi-automated data release - easy access portal with audit trail
- workshops and surveys to bring general public up-to-speed with what is a TRE, feedback about the online portal design
- final report
- broad suppot for social and health care data in reseach
- transparency and accountability were considered important
- broad but conditional support for semi-automated output checking
- the DARE Driver Project on Standardised Architecture for Trusted Research Environments
- "TREs are the future" :sunglasses:
- Too many implementations of TREs - cloud/on-prem, community/commercial
- part of SATRE - open code, open documentation, open communication
- Fortnihtly online cafe for TREs
- Three fold strategy - learn and explore, simplify and specify, guide and sustain
- Research of an architecture / specification , rather than necessarily dictating a solution or product.
- DARE Teleport - Federated Data to Support Team Science
- National TRE of Scotland and Wales - providing plumbing between TREs to allow them to commucate
- Ephemeral Project specific TRE - using existing governance moderls
- IaC to provision a full TRE very quickly - terraform, ansible
- The governance people have "rolled over"
- The value of Federated TREs: The Scottish Safe Haven Network
- "How do you solve a problem like data sharing?"
- Project specific data sharing agreements
- Complexities between data controllers and data processors
- Lengthy legal and information governance reviews
- Differences in risk appetite and local permissions processes.
- The solution...?
- Idetify common data sharing scenarios
- Documenting processes and workflows used for previous projects.
- Demonstrate equivalent information security
- Increase trust between Health Boards and corresponding TREs
- Generic data sharing agreements covering most types of data sharing.
- Link Project TREs to data SDEs
- TREs are projects - we should be bringing the data to the TRE, rather than the TRE to the project
- TREs support the research process - make hypothesis -> publish results
- How can we make a common infrastruture in a public TRE (? may have misunderstood)
- areas to collaborate:
- separate TREs from SDEs.
- Common identity management
- Common research project registration.
- common infrastructure
- SHARE - an architectural structure
- consent and rights embedded
- Opportunities for controlling ML disclosure risks within a TRE
- what is disclosure control? any control to reduce the risk of an individual being identified
- all trained ML models could contain such information
- what controls could be applied to mitigate the risks? legal, procedural and technical
- heads up about expectations
- include hold out data for validation later
- cost for disclosure control
- work with data controllers
- data minimisation and de-identification
- monitor what happens to data in the TRE
- data encoding, overfitting - attack vulnerability
- model vs instance learners
- supporting neuroimaging research within a trusted research environment
- dementia platform uk is a TRE to enable research into dementia
- 19 of 63 datasets contain imaging data which has lead to a big increase in the number of AI/ML projects.
- challenges include storage, computation, data size requirements, model release
- SeRP provide a platform to try and tackle these challenges and allows data access from their own HPC cluster.
- looking into synthetic imaging data, ai/ml model research federation of neuroimaging data
- Using magic to find bugs in your TRE code (aka property based testing for )
- :exploding_head: Testing Matters :exploding_head:
- obvious answers to testing are distracting
- "throwing a toddler at your computer and getting it to smush the keyboard"
- Commercialing technology developed through TRE data access
- Steps to Success
- TRE terms - know your TRE. get your hands on thier terms as early as possible with IP being particularly important
- Confidentiality - make sure your output is protected
- Other agreements and policies - third parties who have an interest in the project, funding agreements, collaboration agreements, etc.
- Maximising generation of key IP - what is the best strategy for commercialising what you "own"
- Records - make sure you have very detailed records of what has been happened, what data was used and by who / for what. exciting things come under close scrutiny.
- A demo of OMECU
- it looks nice :slightly_smiling_face:
- building TREs for Life Sciences

## Lightning Talks (Room 2)

:cry:

###

## Breakout Rooms Session 1

1.[Multi-TRE analysis: challenges, governance requirements, federation](https://hackmd.io/@manics/HyA-LLQ0n/edit?both)
2. +3 [Current state of the art re data linkage/federation/AI&ML&LLM across infrastructures: federation, governance, safe output methods](https://hackmd.io/@manics/Sk1ML8mA3/edit?both)
4. [TRE sustainability and operations](https://hackmd.io/@manics/BklfLU7Rh/edit?both)
5. [Safety and security of Python and R package import into TREs](https://hackmd.io/@manics/SkZGLL702/edit?both)
6. +7 [Community-based efforts and collaborations in public involvement and engagement](https://hackmd.io/@manics/SJzMLU7A2/edit?both)

8. [Cloud vs on-prem TREs: costs, constraints, pros & cons](https://hackmd.io/@manics/rkEzLI7Rh/edit?both)
9. [The role of independent TRE providers in relation to the NHS national and regional SDEs](https://hackmd.io/@manics/SkSzIUQRh/edit?both)
10. [Future governance of the SATRE Specification](https://hackmd.io/@manics/Sksf8LQRn/edit?both)
11. [What are the detractors to standardisation and wSupporting evolution of platforms / servics - budegt and funding
**Cant see the Woods for the TREs**
CHangeing in flight - how to manage updates to existing services / platforms in use
No ability to ranslate research into applied use of TRRE things. Devops models and supporting
Data controllers risk appetite and policys needing alignment against something common
Consentual management for the person : How does the citizen become involved in the TRE across TREs.
Iniertia and reluctance to change their worlds towards a harmonised picture - Budgets and funding to do plus stick.
Purpose is lost as to what a TRE SDE etc is.
No clear definitions of these and separations of their function and facilities.

12. [General discussion](https://hackmd.io/@manics/HkbznLQAh/edit?both)

## Breakout Rooms Session 2
1. [Access and rights metadata to support process choreography and interoperability (e.g. generic researcher applications and authorization)](https://hackmd.io/@manics/rJ5MUUQAh/edit?both)
2. [Addressing data harmonisation between different datasets: do TREs have a role?](https://hackmd.io/@manics/HJFMLU703/edit?both)
3. [Sight unseen: how far can we go with keeping data hidden from users?](https://hackmd.io/@manics/ByuMILXR3/edit?both)
4. [TRE sustainability and operations](https://hackmd.io/@manics/BklfLU7Rh/edit?both)
5. [Safety and security of Python and R package import into TREs](https://hackmd.io/@manics/SkZGLL702/edit?both)
6. [Cloud vs on-prem TREs: costs, constraints, pros & cons](https://hackmd.io/@manics/rkEzLI7Rh/edit?both)
7. [Governance of the UK TRE Community](https://hackmd.io/@manics/rknZ8Im0n/edit?both)
8. [Accrediting people as well as environments… will developing “trusted researchers” allow us to do better/more/quicker data science?](https://hackmd.io/@manics/BypML8QC2/edit?both)
9. [The role of independent TRE providers in relation to the NHS national and regional SDEs](https://hackmd.io/@manics/SkSzIUQRh/edit?both)
10. [General discussion](https://hackmd.io/@manics/HkbznLQAh/edit?both)
11. [What are the detractors to standardisation and working together? How can the reasons for the tension work alongside standardisation?](https://hackmd.io/@manics/Bk0GIImRh/edit?both)
12. [General discussion](https://hackmd.io/@manics/HkbznLQAh/edit?both)