MultiModalHumanLungAgePrediction

This repository provides the code and the supporting methods and results for the paper: "Predicting age from human lung tissue through multi-modal data integration" A Moraes, M Moreno, R Ribeiro, PG Ferreira

International Conference on Discovery Science, 644-658

---

Pipelines

Gene Expression Pipeline

Pipeline for training and test of the gene expression model.

Methylation Pipeline

Pipeline for training and test of the DNA methylation model.

Feature Selection and SMOGN on Methylation data. Application of the SMOGN data pre-processing following Branco et al [1] with the implementation from [2].

Histological images Pipeline

UMAP based on the Haralick Features of 90 Lung samples. No clear separation is found on the age of the individuals based on the Haralick features.

Histological images Pipeline

CNN optimal Parameters. Parameters for the best performing model on the histological image regression.

---

References

1. Branco, P., Torgo, L., Ribeiro, R.P.: Smogn: a pre-processing approach for im- balanced regression. In: First international workshop on learning with imbalanced domains: Theory and applications. pp. 36–50. PMLR (2017)

2. https://pypi.org/project/smogn/, Nick Kunz.