Skip main navigation

Integrating genomic and epidemiological AMR data

introduction to data integration

In the previous step, we saw how comparing sequences to each other can begin to uncover the transmission relationships between isolates. This is the basis of genomic epidemiology. However, just comparing sequences won’t gain us much insight. For this, we need to integrate metadata into our analyses.

What is data integration?

Data integration refers to the process of combining diverse datasets, such as genomic sequences, patient metadata, clinical outcomes, and geographical or temporal information, into a unified framework. This integration enables comprehensive analyses of pathogen evolution, transmission, and resistance patterns. By linking genetic and epidemiological data, researchers can trace outbreaks, identify high-risk clones, and support real-time public health responses more effectively and accurately.

How to integrate different sources of data?

Incorporating different types of data into your analysis can range from simple data column combination in a table to full integration of data into epidemiological models or their equivalents. Below is a table that outline three primary ways to do this:

Approach & example program(s) (click link for training guide) Advantages Disadvantages Best for
Manual data entry (e.g. Excel) Easy to use with gentle learning curve; Advanced use can allow some automation Limited automation options; high chance of making entry mistakes; not very scalable; limited data analysis tools Small datasets where analysis is unlikely to be repeated
Workflows (e.g. Data-Flo) Scalable analysis with no/little coding needed; easy to use visualisations and standard analyses. Can get input from multiple types of data sources, such databases, Excel files of text files. Can be integrated with Microreact Some loss of flexibility compared to programmatic approaches; lacking advanced data analysis tools. Does not work on very large datasets, with more than 10000 records. Range of data sizes with no coding requirements
Programmatically (e.g. R; Python) Versatile platforms for extended data combination and analysis; scalable to very large datasets; version control and high reusability Steep learning curve; visualisations less intuitive than above options Large automated analyses with advanced analysis (e.g. machine learning)

Which of these approaches do you think you may use for your own work? Think about your dataset sizes and automation requirements.

© Wellcome Sanger Institute
This article is from the free online

Antimicrobial Databases and Genotype Prediction: Data Sharing and Analysis

Created by
FutureLearn - Learning For Life