Here we describe the cohorts used for training and testing. An overview of the data, including links to more detailed descriptions of the other data, may be found at https://cmi-x.org/prediction-challenges/
LJI 2025 cohort. In year 1, we collected samples from 40 donors across 5 timepoints. The cohort consisted of 40 participants between the ages of 19 and 65 (Table 1). The study protocol was approved by the Institutional Review Board of the University of Georgia (IRB #20224877). Each participant received a standard Fluzone dose (15 μg of antigen) in the 2025-2026 influenza season (Table 2 shows flu strains present in the vaccine). Samples were collected according to Figure 1.
UGA 9 cohort. In year 0, we performed experiments on the UGA 9 cohort. This cohort consisted of 40 participants between the ages of 19 and 89 (Table 1). The study protocol was approved by the Institutional Review Board of the University of Georgia (IRB #20224877). Each participant received either a standard Fluzone dose (15 μg of antigen for ages < 65) or a high Fluzone dose (60 μg of antigen for ages >= 65) in the 2024-2025 influenza season (Table 2 shows flu strains present in the vaccine). Samples were collected according to Figure 2.
Publicly available data. In addition to the above cohorts, we have collected public data from across numerous resources. Details of these cohorts may be found in the related publications.
As part of the CMI-Flu challenge, we have collected the following information for participants and the investigations they have been involved in:
In the participant_hla.tsv data we provide HLA typing
for participants of the LJI 2025 cohort, as well as SDY314.
LJI 2025 cohort: Genomic DNA from PBMC was typed at the American Society for Histocompatibility and Immunogenetics (ASHI)–accredited Medical Genomics Laboratories, Murdoch University (Western Australia). Locus-specific, sample-indexed (multiplex identifier, MID) primers targeted polymorphic exons for Class I HLA-A, -B, and -C (exons 2–3) and Class II HLA-DQA1/-DQB1 (exons 2–3), -DRB1/3/4/5 (exon 2), and -DPB1 (exon 2). Libraries were prepared using NEBNext Ultra II Library Prep Kit and sequenced on an Illumina MiSeq (v3 600-cycle, 2 × 300 bp). Reads were demultiplexed by MID barcodes, adapter/primer sequences trimmed, quality-filtered, and merged. Alleles were called using the IIID HLA Allele Caller and the IIID HLA Analysis Suite, with the IMGT/HLA Database as the reference.
hla = read_tsv('../datasets/260512/train/participant_hla.tsv')head(hla)## # A tibble: 6 × 5
## subject locus_name allele_1 allele_2 study_accession
## <chr> <chr> <chr> <chr> <chr>
## 1 SUB1829 HLA-A A*24:02 A*24:02 2025LJI
## 2 SUB1829 HLA-B B*40:02 B*40:02 2025LJI
## 3 SUB1829 HLA-C C*03:03 C*03:04 2025LJI
## 4 SUB1829 HLA-DPB1 DPB1*04:02 DPB1*05:01 2025LJI
## 5 SUB1829 HLA-DQA1 DQA1*01:01 DQA1*01:02 2025LJI
## 6 SUB1829 HLA-DQB1 DQB1*05:03 DQB1*06:02 2025LJIprint(paste('Number of donors with HLA typing information:', length(unique(hla$subject))))## [1] "Number of donors with HLA typing information: 119"