VektorPedia

Problem

Plant viruses spread through insects, but the knowledge of which insect carries which virus is scattered across many biodiversity databases. Researchers and plant-disease practitioners had no single place to ask: "which insects are likely vectors for this virus?"

Context

This was my master's thesis at IPB University. The data already existed publicly — in Global Biotic Interactions (GloBI), Wikidata, DBpedia and the NCBI Taxonomy ontology — but in different shapes, identifiers and levels of completeness.

Solution

I built VektorPedia, a search engine backed by an integrated biodiversity knowledge graph. Data from the four sources was ingested and aligned into one graph that captures relationships between insects, viruses and plants. Network analysis on that graph surfaces and scores candidate insect vectors.

Architecture

Technical challenges

• Heterogeneous sources — each knowledge graph uses its own identifiers and vocabulary, so entities had to be aligned before any analysis made sense.
• Signal from structure — interactions are sparse; network analysis was used to infer likely vectors from how species are connected, not just from direct records.
• Research to product — turning a research pipeline into a usable search application within about three months.

Impact

public knowledge graphs integrated

4public knowledge graphs integrated

from design to deployed app

3 mofrom design to deployed app

indexed publication

Scopusindexed publication

The work was published in a Scopus-indexed journal (JISEBI, 2024) and deployed at vektorpedia.ipb.ac.id.

Lessons learned

Knowledge graphs are powerful precisely because they carry context — but most of the effort is in the unglamorous alignment work. A good search experience on top is what makes the research usable by people outside the lab.

Snapshots

Summary of the ideaResearch stagesInteraction graphEmbedded taxonInsect vector scoringDetail information page