Giovanni Parmigiani

The paper identifies genomic scars from alternative end joining and single-strand annealing across multiple cancer types, revealing their association with HR deficiency and local genomic features.

This paper introduces a Bayesian multi-study NMF method to identify common and condition-specific mutational signatures across multiple datasets.

ALADYNOULLI integrates longitudinal health records and genetic data to model disease progression, identifying interpretable disease signatures with cross-population consistency and improving clinical prediction.

This paper performs a meta-analysis of breast cancer risk associated with PALB2 pathogenic variants, finding an overall risk of 12.80% by age 50 and 48.47% by age 80.

The paper extends Mendelian risk prediction models to incorporate variant-specific penetrances, improving cancer risk predictions for BRCA1/2 pathogenic variants.

The paper introduces a hierarchical resampling method for multi-study prediction, showing improved performance in neurochemical sensing applications.

This paper develops an ensemble learning approach to improve genomic classifiers by integrating predictions across batches rather than removing batch effects from merged data, showing better performance in independent validation.

ComBat-Seq adjusts RNA-Seq count data for batch effects using negative binomial regression, improving statistical power and false positive control in differential expression analyses.

This paper shows that molecular subtypes and gene signatures in ovarian cancer are influenced by stromal cell content, suggesting that bulk tissue analysis may be unreliable without accounting for cell admixture.

This paper develops tree-weighting methods for multi-study ensemble learning, showing improved robustness and feature interpretation in genomic datasets.

DeMixT improves transcriptomic deconvolution by estimating component proportions and expression profiles in heterogeneous samples, showing high accuracy in validation and TCGA data.

This study investigates how different sources of heterogeneity affect the accuracy of genomic prediction models across studies, finding that some sources have limited impact while others, like identical generative models, significantly improve consistency.