Biomedical and Health Sciences

Train a CNN-LSTM on overnight pulse-ox traces to estimate sleep-apnea severity, compare models, and sharpen clinical data analysis skills.

Model blood flow in coronary bifurcations, quantify shear stress, and test which plaque sites line up with later imaging.

Use public DNA methylation data to measure epigenetic age acceleration, compare adversity groups, and practice cross-cohort replication.

Model how SARS-CoV-2 epitopes may mimic human self-peptides, then rank cross-reactivity risk with HLA binding and structure similarity.

Use public cancer datasets to find synthetic-lethal KRAS partners and build bioinformatics skill with dependency ranking, co-expression, and validation.

Build a machine-learning classifier for lupus microarray and RNA-seq data, then use SHAP to find genes that separate age of onset.

Measure gait asymmetry from a rear-facing smartphone video, compare it with published norms, and practice pose-estimation analysis for diabetes screening.

Build an ECG classifier that predicts hypertrophic cardiomyopathy, then map model attention to QRS features and test how well it generalizes.

Measure vitamin C loss in cut fruit with iodometric titration, then model how storage temperature changes decay rate and shelf life.

Build a cough-sound classifier from public audio data, learn mel-spectrograms and CNNs, and compare how well it separates respiratory conditions.