Fig. 1
Respiratory Viral DREAM Challenge overview. a Schematic representation of the Respiratory Viral DREAM Challenge workflow. Participants used feedback from evaluation on the leaderboard test set to optimize their T0 and T24 models, and submitted a single model, per timepoint, for final evaluation on the Independent Test Set. b Schematic representing the data provided to participants. 125 subjects were provided as training data, 23 subjects were provided as a leaderboard test set, and 21 subjects from an independent data set were used for final evaluation. c Challenge data come from seven viral exposure trials with sham or one of four different respiratory viruses (H1N1, H3N2, Rhinovirus, and RSV). In each of these trials, healthy volunteers were followed for 7−9 days following controlled nasal exposure to one respiratory virus. Blood was collected and gene expression of peripheral blood was performed 1 day (24−30 h) prior to exposure, immediately prior to exposure and at regular intervals following exposure. Data were split into a training, leaderboard, and independent test set. Outcome data for the leaderboard and independent test set were not provided to the teams, but instead, teams were asked to predict them based on gene expression pre-exposure (T0) or up to 24 h post-exposure (T24). d Histograms and boxplot of the three outcomes by viruses. Symptom data and nasal lavage samples were collected from each subject on a repeated basis over the course of 7−9 days. Viral infection was quantified by measuring the release of viral particles from viral culture or by qRT-PCR (viral shedding). Symptomatic data were collected through self-report on a repeated basis. Symptoms were quantified using a modified Jackson score, which assessed the severity of eight upper respiratory symptoms (runny nose, cough, headache, malaise, myalgia, sneeze, sore throat, and stuffy nose). On the boxplot, the lower whisker, the lower hinge, the mid hinge, the upper hinge and the upper whisker correspond to −1.5× the interquartile (IQR) from the first quartile, the first quartile, the median, the third quartile and 1.5× IQR from the third quartile of the log symptom score, respectively
