TB Biomarker discovery: big data & bioinformatics
Helen Fletcher Associate Professor, Infection & Immunology Department
Director, LSHTM TB Centre Impact of active Genomics • What could happen disease
• What is Transcriptomics happening now
• What makes it Proteomics happen
Metabolomics • What has Impact of happened host environment TB disease studies until 2013
Joosten SA, Fletcher HA, Ottenhoff THM (2013) A Helicopter Perspective on TB Biomarkers: Pathway and Process Based Analysis of Gene Expression Data Provides New Insight into TB Pathogenesis. PLOS ONE 8(9): e73230. doi:10.1371/journal.pone.0073230 http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0073230 • What could Impact of Genomics happen MTB infection
• What is Transcriptomics happening now
• What makes it Proteomics happen
Metabolomics • What has Impact of happened host environment • What could Impact of Genomics happen vaccine
• What is Transcriptomics happening now
• What makes it Proteomics happen
Metabolomics • What has Impact of happened host • Study design – individually matched controls per case (age, gender, birth weight, gestation, environment ethnicity) • Normalize eg. Deconvolution • Discovery Monocyte frequency and phenotype variable in South African infants
Cluster 1 Cluster 2
Differences in; • CD4+ Th1 response • Monocyte frequency • Lymphocyte frequency • Monocyte phenotype (M1/M2) Both responders and non-responders 1 day post MVA85A vaccination
MVA85A (blue) or Candin (red)
Whole blood from infant heel-stick T-cell activation is a correlate of risk in BCG-vaccinated infants
HLA-DR+ CD4+ T-cells BCG-IFN- ELISPOT 400 100
s 300
l l OR 1.828, p = 0.002 OR 0.502, p = 0.013
e
C
c
M
e
B
v
i
l 200
P
f
n
o
o
i
y
l
l
c 10 150
i
n
m
e
/
u
C q 100
F
e
r
S
f
g
o
l 50
1 0 control case control case control case Result significant if Conditional Logistic Regression P<0.05 and FDR<2 Shaded bar indicates medium third of immune response level
BCG Antigen specific IFN-g positive T cells (CD4+ polyfunctional) in peripheral blood associated with reduced risk of disease in BCG vaccinated infants
Fletcher HA et al Nature Communications, 2016 Does MVA85A drive Type I/II IFN and KITAA1 cell967 activation? NUAK2 PDCD4 ICAM1 PDCD7 PDCL3 PSEN1 PILRA PYCARD FPR2 PSMB9 TNFSF10 FPR3 SLC25A4 TXNDC12 RNASE2 SRC XAF1 S100A8 Antiviral, interferon Chemotaxis USP7 Lysosome PDCL3 Apoptosis S100A9 BST2 WDR48 CARD16 TYMP EIF2AK2 chemotaxis lysosome apoptosis BSTIFIH12 C3AR1 ACP2 BCL2A1 IL4I1 EIIRF9F2AK2 CCL2 ARSD BID PSEN1 IFIH1ISG15 CCL3 C1orf85 BLCAP RNASE2 IRF9MX1 CCL3L1 CD68 CASP1 SLC15A3 IPOSGL1R53E CCL4 CTSL1 DNASE2 DWDR48NASE2 MX1RSAD2 CCL4L1 DRAM1 DInRnAM1ate immunity POSTATLR13E CCL7 KIAA1967 IL4I1CFD RIFI44SAD2 CCL8 NUAK2 STIFI35AT1 CXCL10 PSENFCGR1A1 IL4I1 TLR8 ACP2 PDCD4 IFI44TLR8 FPR2 RMARNASE2CO PSEN1 TNF ARSD PDCD7 IFI35TNF FPR3 SLPOCL1R53A3E RNASE2 IFI44L C1orf85 PSEN1 TLR8IFI44L RNASE2 WDR48SERPING1 SLC15A3 IFI27 CD68 PYCARD TNFIFI27 S1IFIT00M3A8 0 2 7 TLR1 WDR48 CTSL1 TNFSF10 IFI44LIFITM3 S1FCGR1B00A9 TLR6 Innate immunity DNASE2 TXNDC12 IFI27FCGR1B TIFIT3YMP IL1A CFD DRAM1 XAF1
IIFIT3FITM3 0 2 7 IFI6 IL6 FCGR1A IL4I1 PDCL3 FCGR1BIFI6 IFIT1 APOBEC3A MARCO Innate immunityPSEN1 CARD16 IFIT3IFIT1 IFIT3 TNFRNASE2
POLR3E 0 2 7 IFI6IFIT3 Host virus interactionIFIT3 PoSLCsi1tive5A3 regulation of NFkB SERPING1 IFIT1IFIT3 IFITM2 LWDR48GALS9 TLR1 IFIT3IFITM2 IFIT2 TInInCaAM2te immunity TLR6 IFIT3IFIT2 CD247 TNFCFD IL1A IFITM2 E4F1 FCGR1A Days post 0 2 7 poly (ADP-ribose) polymeKey:ra foldse fa increasemily in IL6 HNRPA1L IFIT2 PARMARP9CO APOBEC3A MVA85A ICAM1 POLR3E mRNA levels
0 2 7 TNF 0 2 7 PDCL3 SERPING1 Poly (ADP-ribose)Posit ive regulation of NFkB PILRA TLR1 LGALS9 PSMB9 TLR6 0.5 1 2 4 > 5 polymerase TICAM2 SLC25A4 IL1A family TNF SRC IL6 poly (ADP-ribose) polymeraseUSP7 family APOBEC3A PARP9 WDR48 TNF
0 2 7
0 2 7
0 2 7 Challenges • Small sample size - poor statistical Power • Sensitive to type I error (false positive) • Heterogeneity – clinical samples Solutions • False discovery rate (FDR) - if too stringent type 2 error • Larger sample sizes – reduced costs • Normalisation
• Study design - case definition, matched controls
• When is heterogeneity noise and when is it telling you something important? Helen McShane Vivek Naranbhai LSHTM University of Oxford Andrea Zelmer Lisa Stockdale AERAS Satria Prabowo Bernard Landry Peggy Snowden SATVI Bruce McClain Elisa Nemes Tom Evans Michele Tameris Mark Hatherill Tom Scriba Greg Hussey Adam Penn-Nicholson Twitter Hassan Mahomed @Fletcher_Helen Willem Hanekom @LSHTM_TB
Website MVA85A Study team Tb.lshtm.ac.uk All the mothers and babies