Lhasa Limited is excited to announce the anticipated launch of a Federated Learning Hackathon.
Effiris delivers state-of-the-art secondary pharmacology models, where each model learns from the in-house confidential data of pharmaceutical collaborators, while preserving all confidential aspects of the compounds used. As a result, Effiris overcomes one of the main challenges facing model development in drug discovery – accessing and utilising high-quality proprietary data.
This is not the first hackathon in the lifetime of Effiris. In 2019 Lhasa coordinated a proof-of-concept hackathon, to validate the methodology used within Effiris today – since this proof-of-concept exercise Effiris has come a long way – both in terms of software functionality and secondary pharmacology prediction chemical space coverage.
The aim of the Federated Learning Hackathon – expected to launch within Q2 2022 – is to produce one federated Effiris model which outperforms the models trained in-house using only private proprietary data, at each of the participating pharmaceutical organisations. A diverse panel of nine secondary pharmacology endpoints* of high interest to the pharmaceutical industry have been prioritised for the hackathon initiative.
The anticipated benefits to hackathon participants are:
- The opportunity to work with industry peers to build models for vital secondary pharmacology endpoints of clinical interest to the pharmaceutical industry
- An enhanced understanding into how federated learning can progress drug discovery
- Access to Effiris – technology developed by Lhasa alongside the existing Effiris industry consortium – providing an insight into the benefits of Effiris
- Inclusion within the author list on an innovative resulting scientific publication.
A summary of expected hackathon participant activities:
- Attend an online Federated Learning Hackathon kick off meeting within Q2 2022
- Prepare in-house data, in line with a data preparation guide
- Share label files generated from proprietary data with Lhasa (Effiris label files extract knowledge while preserving all confidential aspects of the compounds used)
- Build models from Lhasa consolidated (privacy preserved) label files and send validation results to Lhasa
- Attend a further meeting with Lhasa and all hackathon participants to present and discuss the Federated Learning Hackathon result.
In the below video Anax Oliveira (Director of Science), Laura Johnston (Associate Director of Applied Sciences) and Adrian Fowkes (Principal Scientist) explain more about the Federated Learning Hackathon.
To find out more about this forward thinking, science focused, Federated Learning Hackathon please get in touch.
*The 9 prioritised Federated Learning Hackathon endpoints: Bile salt export pump inhibition, Cyclooxygenase-2 inhibition, GABA-A receptor binding, Glucocorticoid receptor binding, hERG channel inhibition, Muscarinic acetylcholine receptor M1 binding, Phosphodiesterase 4D inhibition, Serotonin 2b (5-HT2b) receptor binding, Tyrosine-protein kinase LCK inhibition.