0.25
0.5
0.75
1.25
1.5
1.75
2
Graph neural networks for computational drug repurposing
Published on Jun 28, 2019157 Views
It can take 15 years and cost $1 billion for a new drug to reach patients as the question of identifying which diseases a new drug could treat is tremendously complex. Diseases are not independent of
Related categories
Chapter list
Part 1: Next-Gen Machine Learning for Network Biology00:00
Two Lectures - 100:18
Preview of Tomorrow’s Lecture00:39
Two Lectures - 203:19
Science crucially depends on scientific instruments03:29
However: Biomedical data present challenges for knowledge discovery03:56
Significant gap04:57
Outline of this Lecture - 105:40
Networks allow for integration of biomedical data06:35
Why Networks? Why Now? - 107:30
Why Networks? Why Now? - 208:15
Why Networks? Why Now? - 308:39
Why Networks? Why Now? - 409:00
Why Networks? Why Now? - 509:15
Many Data are Networks09:27
How to do machine learning on biomedical networks?09:49
Prevailing Deep Models10:27
Biomedical Networks12:04
Why is deep learning on networks hard?12:31
Outline of this Lecture - 213:47
Part 2: Node Embeddings14:13
Embedding Nodes - 114:25
Example: Disease Similarity Network15:50
Setup16:47
Embedding Nodes - 217:25
Embedding Nodes - 318:19
Learning Node Embeddings18:45
Two Key Components19:38
Embedding Methods20:11
Outline of This Section - 121:53
Shallow Node Embeddings22:03
Node Similarity22:39
Node Embeddings24:19
Optimization Task - 125:02
Optimization Task - 226:22
Node Similarity Function Based on Random Walks27:24
Why Random Walks?27:44
Random Walk Optimization - 128:18
Random Walk Optimization - 229:04
Random Walk Optimization - 330:10
Solution: Negative Sampling30:39
Random Walks: Overview31:26
What is the strategy R?31:58
node2vec: Biased Walks32:42
Experiment: Local vs. Global34:06
struc2vec: Structural Similarity37:26
struc2vec: Three Main Steps38:38
struc2vec: Step 139:23
struc2vec: Step 240:54
struc2vec: Step 341:34
struc2vec: Overview42:14
struc2vec: Experiment - 142:52
struc2vec: Experiment - 244:16
Beyond struc2vec: GraphWave45:02
Summary so Far46:18
Outline of This Section - 247:20
Biomedical Applications - 147:27
Biomedical Applications - 247:33
Human Interactome - 148:00
Human Interactome - 248:18
Disease Pathways48:42
Disease Pathways: Task49:15
Disease Pathway Dataset49:56
Pathways: Results51:32
Experimental Setup53:32
Biomedical Applications - 355:37
Protein-Protein Interactions56:21
Network Data56:30
Learning Edge Embeddings - 156:48
Learning Edge Embeddings - 257:31
Experimental Setup - 258:25
PPI Prediction: Results59:42
Biomedical Applications - 401:00:32
Outline of This Section - 301:00:57
Outline of this Lecture - 301:01:03
Part 3: Heterogeneous Networks01:01:28
So far we focused on homogeneous networks!01:01:30
Many Het Nets in Biology01:02:06
Motivating Problem: Prediction of Protein Functions01:03:08
Why is protein function prediction across tissues hard?01:03:59
Motivating Problem: What Does My Protein Do?01:05:50
Multimodal Tissue Networks - 101:07:09
Multimodal Tissue Networks - 201:07:48
Setup: Multimodal Networks01:09:38
Embedding Approach01:11:00
Single-Graph Objective - 101:11:41
Single-Graph Objective - 201:12:47
Summary so Far01:13:32
Recall: Hierarchy of Graphs01:14:04
Cross-Graph Objective - 101:14:34
Cross-Graph Objective - 201:14:57
Embedding Approach: Optimization01:15:43
Embedding Approach: Algorithm01:16:26
Biomedical Application01:18:07
What Does My Protein Do?01:18:17
Data: 107 Tissue Graphs01:18:43
Experimental Setup01:21:08
Results: Protein Function Prediction Across Tissues01:22:25
Case Study: 9 Brain Tissues01:27:14
Multi-Scale Node Embeddings01:27:34
Application: Transfer Learning01:27:50
Outline of this Lecture - 401:30:50
Lecture Resources01:31:25
Two Lectures - 301:33:23