This module enables link prediction in iBKH, i.e., predicting entities that potentially link to the target entity (input).

Method detail

• Step I. Knowledge graph embedding. We use the Deep Graph Library - Knowledge Embedding (DGL-KE) (https://github.com/awslabs/dgl-ke) [1], a Python-based implementation for the advanced KGE algorithms, such as TransE [2], TransR [3], ComplEx [4], and DistMult [5].
• Step II. Link prediction based on the learned embedding vectors of entities and relations.

Usage:

• Specify the Entity Type and Entity Name of the target entity of interest (e.g., Alzheimer’s disease entity [Entity Type = Disease, Entity Name = Alzheimer’s disease ]) in the left sidebar.
• Click Submit to run inference.

It will return lists of drugs, genes, symptoms, and pathways which are most likely to link to the target entity. Such procedure may take minutes.

An example of knowledge discovery for Alzheimer’s disease:

Example

References

1. Zheng D, Song X, Ma C, Tan Z, Ye Z, Dong J, Xiong H, Zhang Z, Karypis G. Dgl-ke: Training knowledge graph embeddings at scale. InProceedings of the 43rd International ACM SIGIR Conference on Research and Development in Information Retrieval 2020 Jul 25 (pp. 739-748).
2. Bordes A, Usunier N, Garcia-Duran A, Weston J, Yakhnenko O. Translating embeddings for modeling multi-relational data. Advances in neural information processing systems. 2013;26.
3. Lin Y, Liu Z, Sun M, Liu Y, Zhu X. Learning entity and relation embeddings for knowledge graph completion. InTwenty-ninth AAAI conference on artificial intelligence 2015 Feb 19.
4. Trouillon T, Welbl J, Riedel S, Gaussier É, Bouchard G. Complex embeddings for simple link prediction. InInternational conference on machine learning 2016 Jun 11 (pp. 2071-2080). PMLR.
5. Yang B, Yih WT, He X, Gao J, Deng L. Embedding entities and relations for learning and inference in knowledge bases. arXiv preprint arXiv:1412.6575. 2014 Dec 20.

This module enables exploration of cohort context based on iBKH.

Given characteristics of a cohort (INPUT), the module will return cohort context in iBKH, i.e., cohort context entities (CCE) such as genes, pathways, drugs, diseases, symptoms, etc. which are potentially associated with the query cohort.

Usage:

• By specifying the Entity Type and Entity Name, the user can add multiple Cohort Entry Entities (CEE) in the left sidebar, describing profiles of the input cohort. A weight w can be specified to each CEE, indicating contribution of the CEE in the query cohort. Value of w is ranging between -1 and 1, where a value close to 1 indicates a positive contribution of the CEE on the cohort, and -1 vice versa. For an example, CEE “Depression” with a 0.75 weight indicates that we observed more “Depression” event in the query cohort compared to healthy controls or general population in an EHR system.
• Click Submit to run inference.

It will return lists of CCEs such as drugs, genes, symptoms, and pathways which are most likely to correlate to the query cohort. Such procedure may take minutes.