Using electronic biology based platform to predict flu

Flu epidemics and potential pandemics pose great challenges to public health institutions, scientists and vaccine producers. Creating right vaccine composition for different parts of the world is not trivial and has been historically very problematic. This often resulted in decrease in vaccinations and reduced trust in public health officials. To improve future protection of population against flu we urgently need new methods for vaccine efficacy prediction and vaccine virus selection.


Introduction
Vaccine effectiveness (VE) against H3N2 viruses is typically lower than VE against influenza H1N1 and/or influenza B viruses. It's not uncommon to see VE of about 30 percent against H3N2 viruses. Furthermore, during the flu season 2017 in Australia, VE of the seasonal flu vaccine was around 10% resulting in record-high numbers of laboratory-confirmed influenza A infections, hospitalizations and deaths 1 . This situation raised concerns that similar could happen in the United States during the flu season 2017/2018, in which H3N2 viruses were predominant. The concerns were based on assumptions that H3N2 viruses in Australia and US were similar, as the classical phylogeny indicated, and because the vaccine composition was identical one could expect comparable levels of VE. Therefore, predicted VE of the flu vaccine in the USA at the beginning of the flu season was around 10% 2 . This prediction was justified and rationalized using the assumption that H3N2 viruses circulating in Australia in the flu season 2016-17 are similar to viruses in the Northern Hemisphere.
Comparison of Australian H3N2 viruses and viruses isolated in the USA at the beginning of the flu season 2017-18, performed using a novel functional phylogenetic tool, demonstrated significant difference between these two groups of viruses 3 . This new information led us to predict that the flu vaccine in US should work in the season 2017-18 just as well as in 2016-17 3 . Our prediction was recently confirmed in the interim CDC estimation of 2017-18 seasonal influenza VE, published and released in February 2018 1 . Moreover, the risk for a (H3N2) associated medically-attended influenza illness was reduced through vaccination by 59% among children aged 6 months through 8 years 1 . In order to assess VE against H3N2 viruses for the next flu season 2018-19 in the United States, we analyzed compatibility between new vaccine virus A/Singapore/INFIMH-16-0019/2016 and H3N2 viruses isolated in 2018 in US. This analysis was performed using the informational spectrum method (ISM) based phylogenetic algorithm, the Informational Spectrum-based Phylogenetic Analysis (ISTREE), which we previously used to assess VE for the flu season 2017-18 3 . This algorithm, which is based on the informational hallmark of proteins that determines their biological function, was previously described in more detail 4 .

Results and discussion
In Figure 1 the ISM-based phylogenetic tree is presented for hemagglutinin HA1 from 68 H3N2 viruses collected in the United States from January to February 2018 and stored in the publicly open database GISAID. As can be seen in this figure, the H3N2 viruses are grouped into two separate clusters and the novel vaccine virus A/Singapore/INFIMH-16-0019/2016 belongs to the small cluster encompassing only 8.8% of analyzed viruses. Previously we showed that 71% of H3N2 viruses isolated in the beginning of the US season 2017-18 were informationally compatible with vaccine virus 3 . This compatibility resulted in good protection against H3N2 viruses in this season 1 . The low informational compatibility between new vaccine virus and H3N2 viruses circulating in US suggests that VE for the next flu season in US could be very low. Of note is that H3N2 virus A/ Hong Kong/4801/2014 in vaccine for the season 2017-2018 better fits US viruses than new vaccine virus A/Singapore/ INFIMH-16-0019/2016. This suggests possibility that VE of the current vaccine could be even higher than that for the new vaccine.
Recently, GISAID released data for H3N2 viruses isolated in Australia in January and February, representing precursors of seasonal flu viruses in Australia in 2018 (Dataset 2. Comparison of these viruses with these collected during the flu season 2017 (Dataset 3) served as a base for prediction of VE during the next flu season in Australia. Unexpectedly, the results of this analysis showed that the predicted responsiveness to new vaccine A/Singapore/INFIMH-16-0019/2016 continuously decreases from May 2017 to February 2018 and increases for the previous vaccine A/Hong Kong/4801/2014 ( Figure 2). This result suggests the low efficacy of the new vaccine against H3N2 viruses in the next flu season in Australia. Monitoring in next moths of H3N2 viruses in Australia will be necessary for confirmation of this prediction.
We propose the "ISM-based phylogenetic algorithm ISTREE analysis" for rapid and accurate analysis of different influenza A viruses that can be used for VE prediction. This is a first report VE prediction prior to flu season using computational analysis. Our prediction has been recently confirmed through laboratory reports released by CDC. Based on current data, we predict low VE for the season 2018/2019 for Australia and US due to vaccine virus selection.

Data availability
Sequence data of the viruses were obtained from the GISAID EpiFlu TM Database. To access the database each individual user should complete the "Registration Form For Individual Users". This form, together with detailed instructions, are available on the website. After submission of the Registration form, the user will receive a password. There are no any other restrictions for the access to GISAID. Conditions of access to, and use of, the GISAID EpiFlu TM Database and Data are defined by the Terms of Use.

Competing interests
No competing interests were disclosed.

Grant information
The author(s) declared that no grants were involved in supporting this work. season using the protein sequence-based clustering algorithm ISTREE. This method has been developed and published by one of the authors several years ago. ISTREE seeks to combine molecular data and provide a clustering with respect to antigenic relatedness of the influenza virus HA. Previous publications provided a retrospective analysis of genetic and antigenic data. Basically, ISTREE is advertised as an alternative to antigenic cartography of influenza viruses using serologic data based on hemagglutination inhibition (HI).
In fact, such algorithms would be of great interest and value in terms of vaccine selection. The prospective statements re appropriate vaccine strains for a future influenza season made by the authors here are quite clear but they are solely based on ISTREE assessments. No flanking/supporting data based on standard HI assays is provided. Assessing the validity of the ISTREE data therefore is not possible. The data may well turn out to be useful but, as it currently stands, they may be totally misleading as well.
Solid antigenic data will be required to validate the potentially useful ISTREE algorithm.

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate? I cannot comment. A qualified statistician is required.
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Partly
No competing interests were disclosed.

Competing Interests:
Reviewer Expertise: Virology, diagnostics, and epidemiology of influenza A I confirm that I have read this submission and believe that I have an appropriate level of expertise to state that I do not consider it to be of an acceptable scientific standard, for reasons outlined above.

Veljko Veljkovic
The major Referee's criticism concerning our study is the lack of serology/virology data to support our approach for making valid predictions. This criticism misses the major point of the study, which is to rapidly release predictions based on in silico screening to allow other scientists to look at it as early as possible and to test it through the flu season. However, as we have mentioned in our paper we used this method previously and if the reviewer would have read the paper carefully, he would have realized that our prediction for the flu season 2017/18 was confirmed several months later by CDC using various serological assays. These data can be found in CDC laboratory reports would have realized that our prediction for the flu season 2017/18 was confirmed several months later by CDC using various serological assays. These data can be found in CDC laboratory reports (see the latest CDC report: ) and present a totally https://www.cdc.gov/flu/weekly/index.htm independent validation of our approach.

Abdel-Satar Arafa
National Laboratory for Veterinary Quality Control on Poultry Production, Animal Health Research Institute, Giza, Egypt The article "Using electronic biology based platform to predict flu vaccine efficacy for 2018/2019" is clearly presented and technically sound enough for publication in F1000Research online. It is well-written and supported by computational well-developed bioinformatics analysis.
I just suggest one observations that should be clarified to enhance the work: - Fig 1 The ISM tree based analysis is not supported by a tool to predict grouping like bootstrapping method to confirm accurate grouping Finally, the article is recommended for publication in its present format.

If applicable, is the statistical analysis and its interpretation appropriate? Not applicable
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
No competing interests were disclosed. Competing Interests: I confirm that I have read this submission and believe that I have an appropriate level of expertise to confirm that it is of an acceptable scientific standard.
The benefits of publishing with F1000Research: Your article is published within days, with no editorial bias You can publish traditional articles, null/negative results, case reports, data notes and more The peer review process is transparent and collaborative Your article is indexed in PubMed after passing peer review Dedicated customer support at every stage For pre-submission enquiries, contact research@f1000.com