Knowledge graphs: A short introduction to the core concepts of biomedical knowledge graphs
Вставка
- Опубліковано 8 лип 2024
- A short introduction to the core concepts of knowledge graphs and their use in biomedicine. The presentation provides a minimal introduction to biomedical knowledge graphs, data sources, heterogeneous networks, graph databases, query languages, and examples of how they are used in biomedicine. You may want to also watch my short introductions to the STRING and DISEASES databases, which are used in the construction of knowledge graphs: • The STRING database: B... • The DISEASES database:...
0:00 Introduction: definition of a knowledge graph and structure of the presentation
0:43 Data sources: ontologies, biomedical databases, manual curation, and text mining
1:42 Heterogeneous networks: definition, challenges, and use of integrative databases
2:29 Data storage: relational/graph databases, Neo4j, and structure of a graph database
3:09 Querying graphs: query languages, querying for nodes and edges, SQL, Cypher, and graph operations
3:57 Biomedical examples: disease gene prediction, drug repurposing, clinical decision making, and importance of open licenses and quality data vs. data storage
To learn more about the Clinical Knowledge Graph, have a look at this preprint: doi.org/10.1101/2020.05.09.08...
The beauty of opensource
Open source and open data licenses, yes! Being able to build upon what others have already done is what makes large knowledge graphs feasible.
Thankuuu u helped me undestand basic concepts a night before metabolic engineering exam
Haha, I always enjoy hearing how people use my videos - I hope your exam went well!
@@larsjuhljensen i kinda bombed a basic question but did the numerical 🥹
@@jatinhemnani6111 sorry to hear, it sucks when you mess up on something basic.
Thank you for creating such useful content. I learnt a lot about network biology from your videos.
I am currently analyzing a time-course transcriptomics dataset and trying to create useful graphs for identifying the perturbed signaling pathways.
Is it possible for you create a video using any example dataset that explains the interpretation of the networks to identify pathways or drug targets?
Glad you find it useful and thanks for the suggestion. What you suggest would definitely require more than one video, though! Analysis of time-course transcriptomics data should be fairly standard; I'm certainly not the expert on it, but it would typically involve running DESeq2 with the right design matrix for your experimental setup. After that you'd have a ranked list of significant genes, and you need to decide a cutoff. This could be done by just choosing some cutoff p-value/FDR, but if at all possible I would want to compare the list to some gold standard to find an optimal cutoff. Once you have your final gene list, I would fire up Cytoscape, use stringApp to fetch a network, import the data and use it to color the nodes by e.g. significance, use the MCL implementation in the clusterMaker2 app to cut the STRING network into modules, and use stringApp enrichment analysis to functionally characterize the clusters. After that, it would mostly be a matter of eye candy, possibly using the Omics Visualizer app to show the actual time course data on the network or stringApp to highlight selected enriched terms or drug targets on the network. This is all covered in one way or another in the hands-on training available from jensenlab.org/training/
I am also planning to make short videos about stringApp and OmicsVisualizer over the coming month.
Hi Prof. Lars, Thank you very much for your videos. I value your efforts.
Could I have kind support if possible. I'm coming from wet lab research and interested in moving to knowledge network and systems biology (Mainly interested in knowledge network). Could you please kindly mentor me where should I start;
1. Any kind of course /curriculum you could recommend
2. What kind of programming languages I should learn
3. Any kind of tips for someone like me willing to self study.
Hi, I cannot recommend any specific courses, since I just do not know any of them well enough. Regarding programming language, the two dominating languages in the field are Python and R. Either will get the job done, but my personal preference would be to go with Python (I think it is a better programming language for learning how to program). In terms of tips, I would try to start analyzing some actual data as soon as possible - it is easy to spend time on learning about more things, but you need to apply what you have learned to real-world data (not toy teaching examples) to get actual experience. Aside from learning to program, you need to learn to work on a Linux command line and to use version control (git).
@@larsjuhljensen Thank you very much for your kind reply. Greatly appreciate it.
Help please
Help with what? You'll need to be a bit more specific than that for anyone to be able to help you ;-)