Timely vaccine strain selection and genomic surveillance improves evolutionary forecast accuracy of seasonal influenza A/H3N2
Curation statements for this article:-
Curated by eLife
eLife Assessment
This study investigated the influence of genomic information and timing of vaccine strain selection on the accuracy of influenza A/H3N2 forecasting. The authors utilised appropriate statistical methods and have provided solid evidence that is an important contribution to the evidence base. While the study addresses a key aspect of public health, the impact is rather limited by its exclusive reliance on predictive methods using genomic information, without incorporating phenotypic data.
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
For the last decade, evolutionary forecasting models have influenced seasonal influenza vaccine design. These models attempt to predict which genetic variants circulating at the time of vaccine strain selection will be dominant 12 months later in the influenza season targeted by vaccination campaign. Forecasting models depend on hemagglutinin (HA) sequences from the WHO’s Global Influenza Surveillance and Response System to identify currently circulating groups of related strains (clades) and estimate clade fitness for forecasts. However, the average lag between collection of a clinical sample and the submission of its sequence to the Global Initiative on Sharing All Influenza Data (GISAID) EpiFlu database is ∼3 months. Submission lags complicate the already difficult 12-month forecasting problem by reducing understanding of current clade frequencies at the time of forecasting. These constraints of a 12-month forecast horizon and 3-month average submission lags create an upper bound on the accuracy of any long-term forecasting model. The global response to the SARS-CoV-2 pandemic revealed that modern vaccine technology like mRNA vaccines can reduce how far we need to forecast into the future to 6 months or less and that expanded support for sequencing can reduce submission lags to GISAID to 1 month on average. To determine whether these recent advances could also improve long-term forecasts for seasonal influenza, we quantified the effects of reducing forecast horizons and submission lags on the accuracy of forecasts for A/H3N2 populations. We found that reducing forecast horizons from 12 months to 6 or 3 months reduced average absolute forecasting errors to 25% and 50% of the 12-month average, respectively. Reducing submission lags provided little improvement to forecasting accuracy but decreased the uncertainty in current clade frequencies by 50%. These results show the potential to substantially improve the accuracy of existing influenza forecasting models by modernizing influenza vaccine development and increasing global sequencing capacity.
Article activity feed
-
eLife Assessment
This study investigated the influence of genomic information and timing of vaccine strain selection on the accuracy of influenza A/H3N2 forecasting. The authors utilised appropriate statistical methods and have provided solid evidence that is an important contribution to the evidence base. While the study addresses a key aspect of public health, the impact is rather limited by its exclusive reliance on predictive methods using genomic information, without incorporating phenotypic data.
-
Reviewer #1 (Public review):
Summary:
In the paper, the authors investigate how the availability of genomic information and the timing of vaccine strain selection influence the accuracy of influenza A/H3N2 forecasting. The manuscript presents three key findings:
(1) Using real and simulated data, the authors demonstrate that shortening the forecasting horizon and reducing submission delays for sharing genomic data improve the accuracy of virus forecasting.
(2) Reducing submission delays also enhances estimates of current clade frequencies.
(3) Shorter forecasting horizons, for example, allowed by the proposed use of "faster" vaccine platforms such as mRNA, resulting in the most significant improvements in forecasting accuracy.
Strengths:
The authors present a robust analysis, using statistical methods based on previously published …
Reviewer #1 (Public review):
Summary:
In the paper, the authors investigate how the availability of genomic information and the timing of vaccine strain selection influence the accuracy of influenza A/H3N2 forecasting. The manuscript presents three key findings:
(1) Using real and simulated data, the authors demonstrate that shortening the forecasting horizon and reducing submission delays for sharing genomic data improve the accuracy of virus forecasting.
(2) Reducing submission delays also enhances estimates of current clade frequencies.
(3) Shorter forecasting horizons, for example, allowed by the proposed use of "faster" vaccine platforms such as mRNA, resulting in the most significant improvements in forecasting accuracy.
Strengths:
The authors present a robust analysis, using statistical methods based on previously published genetic-based techniques to forecast influenza evolution. Optimizing prediction methods is crucial from both scientific and public health perspectives. The use of simulated as well as real genetic data (collected between April 1, 2005, and October 1, 2019) to assess the effects of shorter forecasting horizons and reduced submission delays is valuable and provides a comprehensive dataset. Moreover, the accompanying code is openly available on GitHub and is well-documented.
Weaknesses:
While the study addresses a critical public health issue related to vaccine strain selection and explores potential improvements, its impact is somewhat constrained by its exclusive reliance on predictive methods using genomic information, without incorporating phenotypic data. The analysis remains at a high level, lacking a detailed exploration of factors such as the genetic distance of antigenic sites.
Another limitation is the subsampling of the available dataset, which reduces several tens of thousands of sequences to just 90 sequences per month with even sampling across regions. This approach, possibly due to computational constraints, might overlook potential effects of regional biases in clade distribution that could be significant. The effect of dataset sampling on presented findings remains unexplored. Although the authors acknowledge limitations in their discussion section, the depth of the analysis could be improved to provide a more comprehensive understanding of the underlying dynamics and their effects.
-
Reviewer #2 (Public review):
Summary:
The authors have examined the effects of two parameters that could improve their clade forecasting predictions for A(H3N2) seasonal influenza viruses based solely on analysis of haemagglutinin gene sequences deposited on the GISAID Epiflu database. Sequences were analysed from viruses collected between April 1, 2005 and October 1, 2019. The parameters they investigated were various lag periods (0, 1, 3 months) for sequences to be deposited in GISAID from the time the viruses were sequenced. The second parameter was the time the forecast was accurate over projecting forward (for 3,6,9,12 months). Their conclusion (not surprisingly) was that "the single most valuable intervention we could make to improve forecast accuracy would be to reduce the forecast horizon to 6 months or less through more rapid …
Reviewer #2 (Public review):
Summary:
The authors have examined the effects of two parameters that could improve their clade forecasting predictions for A(H3N2) seasonal influenza viruses based solely on analysis of haemagglutinin gene sequences deposited on the GISAID Epiflu database. Sequences were analysed from viruses collected between April 1, 2005 and October 1, 2019. The parameters they investigated were various lag periods (0, 1, 3 months) for sequences to be deposited in GISAID from the time the viruses were sequenced. The second parameter was the time the forecast was accurate over projecting forward (for 3,6,9,12 months). Their conclusion (not surprisingly) was that "the single most valuable intervention we could make to improve forecast accuracy would be to reduce the forecast horizon to 6 months or less through more rapid vaccine development". This is not practical using conventional influenza vaccine production and regulatory procedures. Nevertheless, this study does identify some practical steps that could improve the accuracy and utility of forecasting such as a few suggested modifications by the authors such as "..... changing the start and end times of our long-term forecasts. We could change our forecasting target from the middle of the next season to the beginning of the season, reducing the forecast horizon from 12 to 9 months.'
Strengths:
The authors are very familiar with the type of forecasting tools used in this analysis (LBI and mutational load models) and the processes used currently for influenza vaccine virus selection by the WHO committees having participated in a number of WHO Influenza Vaccine Consultation meetings for both the Southern and Northern Hemispheres.
Weaknesses:
The conclusion of limiting the forecasting to 6 months would only be achievable from the current influenza vaccine production platforms with mRNA. However, there are no currently approved mRNA influenza vaccines, and mRNA influenza vaccines have also yet to demonstrate their real-world efficacy, longevity, and cost-effectiveness and therefore are only a potential platform for a future influenza vaccine. Hence other avenues to improve the forecasting should be investigated.
While it is inevitable that more influenza HA sequences will become available over time a better understanding of where new influenza variants emerge would enable a higher weighting to be used for those countries rather than giving an equal weighting to all HA sequences.
Also, other groups are considering neuraminidase sequences and how these contribute to the emergence of new or potentially predominant clades.
-
Author response:
Thank you to the reviewers and editors for their positive and constructive comments. Based on this feedback, we can see that we need to clarify that the primary goal of this paper is a test of potential changes in public health policy rather than a test of technical improvements to forecasting models. We briefly summarize the primary goal below to address these public reviews and list our proposed revisions to the manuscript based on reviewer feedback.
All real-time forecasting models contend with 2 major constraints:
(1) How far into the future they have to predict
(2) How rapidly the data used for predictions become available in real time
In the case of evolutionary influenza forecasts, the current values of these constraints are 1) 12 months into the future and 2) an average lag of ~3 months for hemagglutinin (HA) …
Author response:
Thank you to the reviewers and editors for their positive and constructive comments. Based on this feedback, we can see that we need to clarify that the primary goal of this paper is a test of potential changes in public health policy rather than a test of technical improvements to forecasting models. We briefly summarize the primary goal below to address these public reviews and list our proposed revisions to the manuscript based on reviewer feedback.
All real-time forecasting models contend with 2 major constraints:
(1) How far into the future they have to predict
(2) How rapidly the data used for predictions become available in real time
In the case of evolutionary influenza forecasts, the current values of these constraints are 1) 12 months into the future and 2) an average lag of ~3 months for hemagglutinin (HA) sequences to become available after sample collection. Regardless of the predictors we use in these models (genetic or phenotypic), our units of prediction always depend on HA: the HA protein is the primary target of our immunity, HA is the only gene whose composition is determined by the vaccine selection process, and influenza diversity is historically defined by clades in HA phylogenies.
Our primary goal of this study was to understand the relative effect sizes of these two common constraints on forecasting while holding all other variables as constant as possible. With this understanding, we hoped to better inform public health priorities and set realistic expectations for current and future forecasting efforts regardless of the technical specifications of each forecasting model. In other words, the goal of this study was not to optimize prediction methods but to estimate the effects of potential policy changes on forecast accuracy.
We found that reducing how far into the future we need to predict consistently reduced our forecasting error in simulated populations (where we knew the true fitness of each virus) and in natural populations (where we either estimated fitness from genetic predictors or we knew the true fitness of each virus based on its future success). Figure 6 and its first supplemental figure show these effect sizes for natural and simulated populations, respectively, when the future fitness of each virus is known at the time of prediction. By definition, we cannot hope to improve our estimates of viral fitness for these forecasts by using other genetic or phenotypic information.
Figure 6 shows that reducing how far into the future we need to predict from 12 to 6 months improves our forecasting accuracy 3 times as much as reducing the lag between sample collection and HA sequence submission to public databases. The impact of this finding is the confirmation that a faster vaccine development process would improve our forecast accuracy substantially more than faster turnaround between sample collection and sequence submission. If our public health goal is to make better predictions of future influenza populations, then this result indicates that our main priority is to speed up the vaccine development process.
If our public health goal is to better understand the composition of currently circulating influenza populations (the units of our forecasts), then Figure 3 shows that reducing the lag between sample collection and HA sequence submission from ~3 months on average to 1 month on average reduces our uncertainty in current clade frequency estimates by half. This impact is also independent of the predictors we use in our forecasting models and is not lessened by the lack of other genetic or phenotypic information in our analyses.
We realize that neither a 6-month vaccine development process nor a 1-month average sequence submission lag exist yet, but we believe that these are realistic and achievable goals for scientific and public health communities. We also realize that these public health goals are not mutually exclusive. By measuring the effects of these realistic changes to current policy through our forecasting experiments, we hope to inspire and motivate researchers and decision-makers who are empowered to make both of these goals a reality.
Finally, we want to emphasize that the use of phenotypic data in forecasts introduces additional delays caused by the lag between when genetic sequences become available and when serological experiments can be performed. Most WHO influenza collaborating centers use a "sequence-first" approach where they characterize the genetic sequence and use available sequences to prioritize phenotypic experiments with serology. This additional lag in availability of phenotypic data means that a forecasting model based on genetic and phenotypic data will necessarily have a greater lag in data availability than a model based on genetic data only. This lag is important for practical forecasts, too, but because the lag reflects specific characteristics of each collaborating center and not a global policy change, we believe this topic falls outside of the scope of this study.
Based on these public reviews and the private recommendations from reviewers, we plan to make the following revisions to this manuscript.
● Clarify the introduction, discussion, and abstract to emphasize the primary goal for this study to test effects of realistic changes to public health policy and note that this study does not cover improvements to forecasting models. As part of these changes, we will include a rationale for our choice of a genetic-information-only approach rather than a model that integrates phenotypic data. We will also refine Figure 1 to more clearly communicate the two factors we tested in this study.
● Provide a clearer explanation for the subsampling approach we use, include supplemental materials to communicate the geographic and temporal biases that exist in available HA sequence data, and discuss potential effects of different subsampling strategies.
● Evaluate the robustness of our results to different randomly subsampled data. We will perform additional technical replicates of our analysis workflow for natural populations, and summarize the effects of realistic interventions across replicates in a supplemental figure and the main text of the results.
● Investigate time-dependent effects of forecast horizons and submission lags on model accuracy to identify any potential biases in accuracy during specific historical epochs or any seasonal trends in accuracy associated with predicting future populations for the Northern or Southern Hemispheres.
● In the discussion, clarify how reducing submission lags would practically improve the WHO's ability to select vaccine candidate viruses and minimize jargon that currently makes the discussion less accessible to the average reader.
● Investigate how changes in forecast horizons and submission lags change the distance between predicted and observed future populations at antigenic positions (i.e., "epitope" positions) to understand whether we see the same effects with that subset of positions as we see across all HA positions.
-
-
-
