The EMPRES-i genetic module: a novel tool linking epidemiological outbreak information and genetic characteristics of influenza viruses

Abstract

Combining epidemiological information, genetic characterization and geomapping in the analysis of influenza can contribute to a better understanding and description of influenza epidemiology and ecology, including possible virus reassortment events. Furthermore, integration of information such as agroecological farming system characteristics can provide new knowledge on risk factors of influenza emergence and spread. Integrating viral characteristics into an animal disease information system is therefore expected to provide a unique tool to trace-and-track particular virus strains; generate clade distributions and spatiotemporal clusters; screen for distribution of viruses with specific molecular markers; identify potential risk factors; and analyze or map viral characteristics related to vaccines used for control and/or prevention. For this purpose, a genetic module was developed within EMPRES-i (FAO's global animal disease information system) linking epidemiological information from influenza events with virus characteristics and enabling combined analysis. An algorithm was developed to act as the interface between EMPRES-i disease event data and publicly available influenza virus sequences in OpenfluDB. This algorithm automatically computes potential links between outbreak event and sequences, which are subsequently manually validated by experts. Subsequently, other virus characteristics such as antiviral resistance can then be associated to outbreak data. To visualize such characteristics on a geographic map, shape files with virus characteristics to overlay on other EMPRES-i map layers (e.g. animal densities) can be generated. The genetic module allows export of associated epidemiological and sequence data for further analysis. FAO has made this tool available for scientists and policy makers. Contributions are expected from users to improve and validate the number of linked influenza events and isolate information as well as the quality of information. Possibilities to interconnect with other influenza sequence databases or to expand the genetic module to other viral diseases (e.g. foot and mouth disease) are being explored. Database OpenfluDB URL: http://openflu.vital-it.ch Database EMPRES-i URL: http://EMPRES-i.fao.org/.

Figure 1.

Linking EMPRES-i and OpenFluDB has implied the development of web services (terminology—query) for data comparison, the creation of an algorithm as the interface between both databases and enable the exchange of data (no transfer).

Figure 2.

Flowchart showing steps to synchronize the linkage between EMPRES-i disease events to OpenFluDB isolates. The procedure has four steps and is based on web services (WS) to exchange information between both databases. (1) OpenFluDB gets a list of disease events described by host, location and date from EMPRES-i. 2) A procedure runs on OpenFluDB to assign scores to computed links between EMPRES-i disease events and OpenFluDB isolates. The computed links with scores are then communicated to EMPRES-i. 3) FAO experts validate the computed links. 4) The validated links are back-communicated to OpenFluDB to be shown as URLs to EMPRES-i on the corresponding OpenFluDB isolate record pages. EMPRES-i queries OpenFluDB to get genetic information about linked isolate(s). All these steps are run daily.

Figure 3.

Example of epidemiological Record of Disease Event in EMPRES-i including validated links to genetic data: (A) with one linked isolate, and (B) with several linked isolates (validated and proposed), several sequence records and identified key mutations. Selecting the SIB icon or individual links under ‘Segments Sequences’ for GenBank will open a new browser window or link to display the sequence record.

Figure 4.

Evaluation of automatic matching by statistical comparison of all computed versus validated or proposed link scores. (A) ROC curve of true positive vs. false positive rate after optimizing the scoring system with a linear discriminant analysis. (B) Log frequency plot of 323 234 computed link scores. Each event-isolate link satisfies the three minimal linkage criteria: same country, same animal class, known year and month of the collect date, date delay ±18 days. Each score is a weighted sum of 3 subscores showing quality of match for species, geographical location and date, divided by the rank of this disease event among all other disease events linked with the same isolate. (C) Distribution of ranks of isolates in the expert approved (validated) links. Eighty-five percent of validated links are between an event and the top-ranked isolate. (D) Distribution of validated link scores. The validated links have significantly higher scores than all computed links (mean scores of 28.5, respectively 6.9. P-value of 2.2^e-16 with a Wilcoxon rank test).