Virome-wide detection of natural infection events and the associated antibody dynamics using longitudinal highly-multiplexed serology

Figures

• 

  Fig. 1. Virus-specific events inferred from co-ordinated temporal changes in highly-multiplexed antibody reactivity data. a Conceptual figure showing how viral events—corresponding to infections or reactivations—appear in a longitudinal series of multiplexed antibody reactivity data. b, c Example plots showing the transformation of peptide-level signals (b) into species-level signals (c) from two consecutive timepoints from an individual participant, separated by 180 days: (b) shows the magnitude of reactivity at each timepoint against 15,000 30mer peptides in the HV2T human virome library, each peptide is represented by a dot whose color and shape reflect its species of origin, while (c) shows the change in species-level reactivity to each of the 80 virus species covered in HV2T, expressed as an Enrichment Score (x-axis: positive=increasing over time, negative = decreasing over time) and adjusted p value (y-axis). The data shown in (c) were calculated using the algorithm described in (d). Species with adjusted p < 0.01 are highlighted, using symbols matching (b). d To detect viral events, our algorithm considers each pair of consecutive samples (“a” and “b”) in a longitudinal series and…

• 

  Fig. 2. Detection of events across the virome in a cohort spanning ~100 total person-years. a For all 65 participants (columns) in the ACS cohort for whom data on 4 × 6-monthly timepoints were available, PSEA analysis was applied to each of the 3 pairs of consecutive timepoints across all 80 viral species represented in the HV2T library. Shown in rows are the 48 species for which ≥1 significant viral event (adjusted p value < 0.01) was observed across the cohort. Significant events are colored by the interval(s) in which they were detected. b Example peptide-level scatterplots for 6 of the (659 total) viral events identified in (a). Each plot shows signals from the 15,000 HV2T peptides (1 dot per peptide) at the 2 timepoints spanning the focal event (x and y-axes), with all peptides from the focal species highlighted in red. c Shown for each of the example events in (b) is the baseline-subtracted reactivity signal (y-axis) across all 4 timepoints (x-axis) of the leading-edge peptides from the focal virus (each peptide represented as a gray line). Black arrows highlight the interval in which PSEA-detected the significant VAE (from which the leading-edge peptides were identified), and the…

• 

  Fig. 3. Seasonality analysis reveals epidemic waves for a subset of viral species. a For each species, the viral events detected across the cohort described in Fig. 2 were localized to one of the sampling 6-monthly intervals, and seasonal effects were quantified by comparing the observed distributions against a one-sided multinomial model in which events were distributed across intervals with equal probability. Shown is a quantile-quantile plot of unadjusted −log10(probabilities) for each of 25 species for which ≥5 total events were observed across the cohort. The five species with FDR-adjusted p values < 0.1 are highlighted and colored according to the interval in which their event count is maximum. b For the five significant species identified in (a), and Rhinovirus B as a negative control, profiles of cohort-wide reactivity dynamics were generated by assigning to each day in the sampling period (Jul 12 2006–May 26 2008) scores representing the number of participants for whom reactivity was significantly increasing (red) or decreasing (blue), normalized to the total number of participants with samples spanning that day. Increasing/decreasing reactivity assignments were based on…

• 

  Fig. 4. Epitope-level analysis resolves Influenza A virus and Enterovirus D epidemic waves to the sub-species level. a Leading-edge peptides from the events detected in Fig. 2 for Influenza A virus (upper) and Enterovirus D (lower) were mapped onto representative viral proteomes of the dominant respective circulating subtypes (Influenza A virus—H1N1, H3N2; Enterovirus D—D68, D70, D94). Shown in the upper rows of each panel are multiple sequence alignments for the reactive proteins of each virus, colored at each position by the fraction of total events whose leading-edge peptides contain the 7mer at that position for each subtype. The lower row of each panel shows average amino acid identity between the subtype representatives across a sliding window of 9 amino acids centered at each amino acid position in the alignment. The dotted box in the Enterovirus D panel shows the position of the surface protein VP1, and the star indicates a frequently-recognized variable region containing D68-specific reactivity. b To resolve the subtype identity of each Influenza A virus and Enterovirus D event, sets of subtype-specific 7mers were quantified for each virus/subtype—both in the total library (top…

• 

  Fig. 5. Correlated dynamics of inflammatory proteins, symptoms and anti-viral antibodies measured in longitudinal dried blood spot sample series. a 379 serial dried blood spot samples collected from a participant over a ~7 year span were evaluated for abundance of 4 inflammatory proteins—C-reactive protein (CRP), Serum Amyloid A (SAA), Myeloperoxidase (MPO), and total Immunoglobulin M (IgM)—using the SISCAPA assay (upper four panels). Cases where measurements exceeded a threshold (dashed red horizontal lines) at ≥2 consecutive timepoints are highlighted (red dots). A subset of 23 samples across the timespan were also evaluated for anti-viral antibody abundance using PepSeq (lower four panels) in which PSEA detected four significant VAEs (asterisks and vertical shading). The red ‡ indicates the timing of an antibiotic-responsive pneumonia. For each VAE, medians of the leading-edge peptide Z-scores are plotted across all 23 timepoints. Horizontal lines show the median Z-scores of leading-edge peptides at the timepoint immediately prior to the corresponding VAE, and the overlaid asterisks indicate subsequent timepoints whose leading-edge peptide signals are significantly elevated over this…