We used Random Network Distilation model (code snippet) for our final submission. Please refer to the paper to learn more about the method.
In vitro cellular experimentation with genetic interventions, using for example CRISPR technologies, is an essential step in early-stage drug discovery and target validation that serves to assess initial hypotheses about causal associations between biological mechanisms and disease pathologies. With billions of potential hypotheses to test, the experimental design space for in vitro genetic experiments is extremely vast, and the available experimental capacity - even at the largest research institutions in the world - pales in relation to the size of this biological hypothesis space.
GeneDisco (published at ICLR-22) is a benchmark suite for evaluating active learning algorithms for experimental design in drug discovery. GeneDisco contains a curated set of multiple publicly available experimental data sets as well as open-source i mplementations of state-of-the-art active learning policies for experimental design and exploration.
Please note that it is possible to open genedisco-pycd/genedisco/visualization/visualization_2.ipynb, even though it's too large to view it on github. To do that, open the notebook as a raw file, copy its content, and save it on your local machine as an .ipynb file. The notebook contains some of our latest visualization comparisons that we used to choose the best performing acquisition function among those we tried.
GeneDisco benchmark is licensed under Apache License.
Contributions from pycd team are licensed under MIT License:
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/core_set.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/core_set2.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/core_setUMAP.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/ensemble_rnd.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/rnd.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/rnd_05.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/uncertainty_acquisition.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/uncertainty_acquisition_03.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/uncertainty_acquisition_05.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/uncertainty_acquisition_07.py
genedisco-pycd/genedisco/active_learning_methods/acquisition_functions/uncertainty_acquisition_10.py
genedisco-pycd/genedisco/visualization/visualization_2.ipynb
genedisco-pycd/genedisco/visualization/viz.py
genedisco-pycd/genedisco/visualization/viz_utils.py
- Panagiotis Tigas (@ptigas)
- Yashas Annadani (@yashasannadani)
- Chris Emezue (@ChrisEmezue)
- Daria Yasafova (@DariaYasafova)