|
|
Post by ACM WAP Moderator on Apr 3, 2021 13:23:41 GMT 8
Title: Graph Representation Learning for Drug Discovery
Author: Dr. Ming Zhang
|
|
|
|
Post by ACM WAP Moderator on Apr 7, 2021 9:55:56 GMT 8
Question #1. I would just like to ask what tools were used from the two papers discussed to replicate the experiments and if there is a need to get opinion from the biomedical field to evaluate the results.
|
|
|
|
Post by Ming Zhang on Apr 7, 2021 10:30:02 GMT 8
For ICLR and ICML papers, the proposed models, as well as all baselines, are implemented using the popular deep learning framework, i.e., Pytorch. The official code is also available online deepgraphlearning.github.io/torchdrug-site/ . The second concern is about the necessity of evaluating the results from a biomedical perspective. This is a good point. The answer is yes. Indeed, evaluation metrics used by the machine learning community may be biased toward a specific property and some results are even ridiculous. The field of drug discovery is multi-discipline and we encourage researchers from all backgrounds to collaborate on this promising field. |
|
|
|
Post by mzhang on Apr 7, 2021 10:57:30 GMT 8
1. Chence Shi*, Minkai Xu*, Zhaocheng Zhu, Weinan Zhang, Ming Zhang, Jian Tang. GraphAF: a Flow-based Autoregressive Model for Molecular Graph Generation. ICLR 2020. (* Equal Contribution) arxiv.org/pdf/2001.09382.pdf 2. Chence Chence Shi, Minkai Xu, Hongyu Guo, Ming Zhang, Jian TangA Graph to Graphs Framework for Retrosynthesis Prediction, ICML 2020. proceedings.icml.cc/static/paper_files/icml/2020/4152-Paper.pdf 3. The official code will be available soon |
|
