Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2)....
Transcript of Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2)....
![Page 1: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/1.jpg)
1
Modeling the Antigenic Evolution of Influenza Viruses
from Sequences Taijiao JiangTaijiao Jiang
Center of Systems Medicine, Chinese Academy of Medical Sciences
Suzhou Institute of Systems Medicine
October 8-10, 2015. Sino-German Symposium, Beijing
![Page 2: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/2.jpg)
2
Overview of the talkI. Backgrounds:
Influenza virus, antigenic evolution and vaccination.
II. Modeling influenza virus antigenic evolution Methods for predicting antigenic variants:
1). EADpred2). Feature-based Naive Bayes model
Methods for predicting antigenic evolution patterns1). Genomic co-occurrence network2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.
III. Application of PREDAC to recommend vaccine strains
for seasonal influenza viruses.
![Page 3: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/3.jpg)
3
Overview of the talkI. Backgrounds:
Influenza virus, antigenic evolution and vaccination.
II. Modeling influenza virus antigenic evolution Methods for predicting antigenic variants:
1). EADpred2). Feature-based Naive Bayes model
Methods for predicting antigenic evolution patterns1). Genomic co-occurrence network2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.
III. Application of PREDAC to recommend vaccine strains
for seasonal influenza viruses.
![Page 4: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/4.jpg)
The impact of influenza virus on public health
Pandemic flu viruses: up to tens of millions deathsSeasonal flu viruses: infecting 10-15% human population each
year, up to ¼ million deaths (H1N1, H3N2 and B).The newly emerging flu viruses: occasional infections with high
rate of death, causing social panic.
Influenza virus A subtypes and its history1918 Spanish Flu
![Page 5: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/5.jpg)
5
Harris et al (2006). Influenza virus pleiomorphy characterized by cryoelectron tomography. PNAS. 103, 19123-19127.
What does a flu virus look like?
![Page 6: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/6.jpg)
6
Influenza genome and proteomeGenetic composition of a flu virus
Negative single-stranded RNA virus.
8 RNA segments encoding at least 12 proteins: NP, PA, PB1 and PB2 code for viral replication machineryNS for non-structural proteins 1 and 2 (NS-1, 2), MP for matrix proteins 1 and 2 (M-1 and 2).HA for hemagglutinin NA for neuraminidase
The surface glycoprotein, HA, controls viral entry into the cells and is the major antigenic target of the host immune responses.
http://micro.magnet.fsu.edu/cells/viruses/influenzavirus.html
![Page 7: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/7.jpg)
7
HA structure
Sauter NK, et al. Biochemistry,1992,31:9609-9621Vishal Gupta, et al. Vaccine,2004,24:3881-3888
Antigenic regions Receptor-binding region
![Page 8: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/8.jpg)
8
Vaccination: the most effective way to control influenza infection in humans
Vaccination Flu seasons:
North Hemisphere: October - April next year.
In mid-Feb of every year, WHO will recommend trivalent vaccine composition: A/H1N1, A/H3N2, and B.
Grown in eggsInactivated virus or live attenuated virusInjection (Flu shot)
![Page 9: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/9.jpg)
Effective vaccination requires that the antigenic property of vaccine strain should
match that of circulating strains.
The critical step in the vaccine strategy is to select the vaccine strains that match future circulating strains in antigenicity .
![Page 10: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/10.jpg)
WHO has set up a Global Influenza Surveillance Network (GISN)
135 National influenza centers from 105 countries6 WHO Collaborating centers(WHOCC: USA, UK, Japan, Australia and China)12 reference labs for H5 viruses
Perform antigenic analysis and HA sequencing
Assess epidemiologic behavior
Identify potential circulating strains
Recommend vaccine strains
![Page 11: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/11.jpg)
11
Current situation: the frequent vaccine mismatches
Our analysis on the vaccine matches from 2002 to 2009.
Du et al. Nature Communications 2012.
![Page 12: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/12.jpg)
12
Influenza virus is highly mutable
Genetic drift to antigenic drift
Co-infection Reassorted viruses
Genetic shift to antigenic shift
1. Gene mutation
2. Gene reassortment
Flu virus mainly uses two strategies to change its antigenicity to escape host immune surveillance.
![Page 13: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/13.jpg)
HI Titer
vaccine strain Circulating strain
Similar Antigenicity Different Antigenicity
The experimental method for flu virus antigenicity analysis: Hemagglutination inhibition (HI) assay
Antigenic variants detectionVaccine strain recommendation
DelayDelay
Disadvantages of HI assay :Labor-intensive & time-consuming
Not consistent & low sensitivity
![Page 14: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/14.jpg)
14
Sequence-based approaches have become important alternatives to assist vaccine strain recommendation
sequence
Identify influenza antigenic variants and recommend vaccine strains
At present, sequencing of HA or even whole viral genome has been a routine work in influenza surveillance.
![Page 15: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/15.jpg)
15
Overview of the talkI. Backgrounds:
Influenza virus, antigenic evolution and vaccination.
II. Modeling influenza virus antigenic evolution Methods for predicting antigenic variants:
1). EADpred2). Feature-based Naive Bayes model
Methods for predicting antigenic evolution patterns1). Genomic co-occurrence network2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.
III. Application of PREDAC to recommend vaccine strains
for seasonal influenza viruses.
![Page 16: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/16.jpg)
16
WHO relies on the two critical steps in vaccine strain recommendation
Step 1. Monitor antigenic variants across the world.
Step 2. Assess their potential dominance in the coming season.
Can we use only sequence information for the mission of the two steps?
![Page 17: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/17.jpg)
The key question 1: how to accurately predict antigenic variants from HA
sequences?
![Page 18: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/18.jpg)
18
Previous models to predict antigenic relationships
Wilson et al. 1990, Annu Rev Immunol 8, 737Lee et al. 2004, Emerg Infect Dis 10, 1385Munoz et al. 2005, Vaccine 23, 1114Gupta et al. 2006, Vaccine 24, 3881Lee et al. 2007, Vaccine 25, 8133Liao et al. 2008, Bioinformatics 24(4), 505Huang et al. 2009, BMC Bioinformatics 10 (Sup 1), S41Lees et al. 2010, Bioinformatics, Epub ahead of print
Either countingamino acid changes on HA, particularly those in the epitope regions.
![Page 19: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/19.jpg)
19
Previous models to predict antigenic relationships
Liao et al. 2008, Bioinformatics 24(4), 505Huang et al. 2009, BMC Bioinformatics 10 (Sup 1), S41Lees et al. 2010, Bioinformatics, Epub ahead of print
Or considering the specific amino acid changes at certain positions on HA observed during influenza evolution (site-dependent models).
![Page 20: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/20.jpg)
20
The motivation of our modeling work
AA Change == Antigenic ChangeInfluenza virus may explore new amino acid changes even different amino-acid positions
during influenza evolution!
So,we are interested in developing a more general
computational model to predict influenza antigenic relationships.
![Page 21: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/21.jpg)
21
We view the antigenic changes as the changes of the interactions between the HA and host antibodies.
Based on HA sequence-structure modeling, we developed two models:
1. EADpred (Epitope-based Antigenic Distance Prediction)
2. Machine learning-based model
![Page 22: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/22.jpg)
EADpred: Epitope-based Antigenic Distance Prediction
Antigenic epitopes as basic structural units to mediate HA-antibody interactions
Physicochemical rules underlying HA-antibody interactions
![Page 23: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/23.jpg)
Four steps of the EADpred
Step 1: Identify the antigenic epitopes
Step 2: Transform amino acid changes in each antigenic epitope into the changes of physicochemical properties
Step 3: Integrate the contributions of all derived antigenic epitopes into a single predictor.
Step 4: Parameterize and assess the model
![Page 24: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/24.jpg)
24
EADpred achieves a good accuracy in quantifying antigenic distances between different viral strains
Testing data: PCC is 0.79
0.7991.380.886.9Epitope-
Based (EADpred)
0.7585.480.883.5Site-dependent*
Coeffi-cient
Speci-ficity
Sensi-tivity
Agree-mentMethod
* Liao et al. 2008, Bioinformatics 24(4), 505
![Page 25: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/25.jpg)
Feature-based Naive Bayes model is a more generalized model to predict antigenic variation
Du et al. Nature communications (2012)
![Page 26: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/26.jpg)
Retrospective Testing shows that feature-based NB model can model antigenic
relationship with high accuracy.
![Page 27: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/27.jpg)
27
WHO relies on the two critical steps in vaccine strain recommendation
Step 1. Monitor antigenic variants across the world.
Step 2. Assess their potential dominance in the coming season.
![Page 28: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/28.jpg)
The key question 2: how to accurately model influenza antigenic patterns for
vaccine recommendation?
![Page 29: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/29.jpg)
Continuous genetic changes Discontinuous antigenic drift
Antigenic cluster
Smith et al. Science 2004.
253 H3N2 viruses during 1968-2002Cluster-wise evolution of influenza antigenicity
HA sequencing HI assay + computational clustering
![Page 30: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/30.jpg)
30
Conventional method to analyze the patterns of influenza evolution : Phylogenetic tree analysis
Trunk-like patternSide lineages Phylogenetic clades
Bush RM, et al. Science, 1999,286:1921-1925
![Page 31: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/31.jpg)
Phylogenetic clades ≠ Antigenic clusters
New computational models are needed to capture the antigenic
evolutionary patterns!!!
![Page 32: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/32.jpg)
32
Based on whole genome information, we have developed a network model to
understand influenza evolution. We developed a network model to capture the co-evolutionary
signals across the whole viral genomes. We used the network to describe human influenza evolutionary
patterns.
![Page 33: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/33.jpg)
Construction of genomic co-occurrence networks
Step 3. Connect the co-occurring nucleotide pairs for the viruses involved
Step 1. Align each of the eight gene segments
Step 2. Compute the co-occurring nucleotide pairs
Each virus was represented as a nucleotide co-occurring network
![Page 34: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/34.jpg)
The structural changes in co-occurrence network rather than the sequence changes capture the antigenic changes of human H3N2 virus
(Du et al., Genome Research 2008)
Contiueous AA changes cannot reflect the clusterwise antigenic changes.
The topology changes of the network are not continueous, capturing the clusterwise antigenic changes of H3N2 viruses.
1D sequence 3D co-occurrence network
![Page 35: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/35.jpg)
Cooperative mutations drive flu antigenic variation.
Exposed side Buried side
Red: Network-captured amino acid changes. Yellow: Other mutations
Antigenic structures of the major antigen HA
![Page 36: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/36.jpg)
Predicting antigenic clusters from HA sequences only
HA
Sequences
Antigenic C
lusters
?
![Page 37: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/37.jpg)
PREDAC: PREDictor of Antigenic Clusters
Smith et al. 2004, Science.Du et al. Nature Communications 2012.
![Page 38: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/38.jpg)
PREDAC vividly delineates the antigenic evolution of H3N2 viruses as the replacement of antigenic clusters
1. ~1100 influenza A (H3N2) viruses isolated between 1968 and 2009.
2. 17/20 dominant antigenic clusters, and each persisted for one to five seasons.
![Page 39: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/39.jpg)
Transmission of H3N2 virus in North and South China
Northern China
Sothern China
![Page 40: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/40.jpg)
40
Overview of the talkI. Backgrounds:
Influenza virus, antigenic evolution and vaccination.
II. Modeling influenza virus antigenic evolution Methods for predicting antigenic variants:
1). EADpred2). Feature-based Naive Bayes model
Methods for predicting antigenic evolution patterns1). Genomic co-occurrence network2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.
III. Application of PREDAC to recommend vaccine strains
for seasonal flu.
![Page 41: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/41.jpg)
The automatic vaccine strain selection
Patent applied.
New Antigenic clusterPercentage >=10%
![Page 42: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/42.jpg)
Comparison with WHO-recommendations
Based on 7 seasons during 2002-2009.
WHO: 2/7 for China;4/7 for North America and Europe
Our accuracy: 6/7 for China;6/7 for North America and Europe
XX
X X X X XX X X
1,Patent obtained。
2,China CDC is using our method in selecting candiate
vaccine strains
3,Nature Communcations, 2012.
![Page 43: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/43.jpg)
XX
X X X X XX X X
![Page 44: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/44.jpg)
Google Flu Trends estimate(black) US CDC data (red)
US Flu activity
Internet-based flu activity predictor
Ginsberg, J. et al. 2009. . Nature 457.
Drawbacks:1. Can not determine which type or subtype of flu virus in circulation?2. Can not determine whether the virus has been mutated?
Therefore, the internet-based approach can not be used to recommend vaccine strains and to select drugs for treating the flu.
![Page 45: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/45.jpg)
Integrated approach: combine internet big data and gene big data for fighting flu
Zou et al. Chinese Sci. Bullet. 2015
![Page 46: Modeling the Antigenic Evolution of Influenza Viruses from ... · Genomic co-occurrence network 2). PREDAC: machine learning-based PREDiction of influenza virus Antigenic Clusters.](https://reader034.fdocuments.us/reader034/viewer/2022052007/601bf90103bf905b73026b48/html5/thumbnails/46.jpg)
46
Acknowledgements
Members in Jiang labAiping WuYousong PengLizong DengYang CaoXiangjun Du
…….
Professors Yuelong Shu (China CDC) Fu Gao (Inst. Microbio, CAS) Daxin Peng (Yangzhou Univ)
Genhong Cheng (UCLA) Jianzhu Chen (MIT)
......
Funding:
Project “973” ; National Key Project for Infectious diseases
NSFC (Outstanding Youth).
Thank you for attention