Table 3. CMI-PB invited prediction challenge methods.

The 25 team submissions were categorized according to their underlying methodology. Additional method characterizations can be found in Supplemental Note 1.

Team	TEAM ID	Synopsis
Sparse linear regression
1	team_52	ImputePCA and training and prediction were made using elastic net regression
2	team9_3	Features were reconstructed using JIVE multi-omics integration, training data consist of 2020 + 2021 datasets, All four assays and subject information was used, training and prediction was done using ElasticNet
3	team9_4	Features were reconstructed using JIVE multi-omics integration, training data consist of 2020 + 2021 datasets, All four assays and subject information was used, training and prediction was done using ElasticNet CV
4	team47	SuperLearner Ensemble
5	team48_1	Features were using MOFA multi-omics integration and final features were handpciked instead of solely relying on LASSO regression, training data consist of 2021 datasets, All four assays and subject information was used, training and prediction was done using LASSO regression
6	team48_2	Features were using MOFA multi-omics integration and final features were handpciked instead of solely relying on LASSO regression, training data consist of 2020 + 2021 datasets, All four assays and subject information was used, training and prediction was done using LASSO regression
7	team5	Establishing purpose-built models using multiple co-inertia analysis, features consist of four omics, baseline values of tasks and
8	team6	Ensemble approach using SPEAR-constructed supervised multi-omics factors with demographic data
9	team9_2	Multi-omics Integration with JIVE and Lasso
10	team_40	Different regression models on multi-omics data using features from the baseline (day 0)
11	team25	Semi-manual feature selection learned between the 2020↔2021 datasets, followed by linear regression
12	team9_1	Multi-omics Integration with JIVE and Basic Linear Regression
13	team49	Dimension reduction through Multiple Co-inertia analysis and modeled with Linear mixed effects
14	team32	Semi-manual feature selection followed by dimensionality reduction and residual from baseline prediction
15	team50	Semi-manual feature selection followed by dimensionality reduction and residual from baseline prediction
Nonlinear regression (regression trees)
16	team_38	Catboost Regression model trained on 2020 training cohort
21	team_53	Catboost Regression model trained on 2021 training cohort
18	team_54	Catboost Regression model trained on 2020+2021 training cohort
19	team45	Model comparison to determine the best algorithm; Manual feature selection; Random forest regression
20	team46	Block forest regression
21	team51	Random forest classifier to simulate training individuals, XGboost to determine final ranking
22	team55	DecisionTree and Random Forest Regressor
Others
23	team30	Fully Connected 2-layer neural network with imputation
24	team34	AutoML based on the most predictive assay or clinical data (trained on 2020 and tested on 2021)
25	team34	AutoML based on the most predictive assay or clinical data (trained on 2020 and tested on 2021)
Control models
26		Use age of study subject as predictor
27		Utilize baseline pre-vaccination state of a task as predictor