Y. Louzoun

Prediction of KIR3DL1/Human Leukocyte Antigen binding

The paper builds a machine learning model to predict KIR3DL1 binding to HLA class I allotypes, showing that binding is influenced by multiple non-additive regions beyond the classical Bw4 motif.

Counting Is Almost All You Need

This paper shows that counting T-cell receptor repertoires outperforms existing methods in detecting CMV and HLA alleles, with an attention model and a simplified graph approach improving accuracy.

Progesterone supplementation in mice leads to microbiome alterations and weight gain in a sex-specific manner

This study shows progesterone increases weight in female mice but not males, with microbiome changes in females only, suggesting sex-specific microbiota involvement in progesterone effects.

MIPMLP - Microbiome Pre-processing Machine Learning Pipeline.

The paper develops a preprocessing pipeline for 16S sequencing data to improve machine learning classification accuracy by optimizing feature representation and dimensionality reduction.

Projection of gut microbiome pre and post-bariatric surgery to predict surgery outcome

The paper projects gut microbiome data pre and post-bariatric surgery to predict BMI loss, identifying distinct microbiome profiles associated with different outcomes.

Autoencoder based local T cell repertoire density can be used to classify samples and T cell receptors

The paper introduces ELATE, an autoencoder-based method to estimate local T cell repertoire density, enabling classification of samples and prediction of TCR-peptide binding.

CDR3 and V- genes show distinct reconstitution patterns in T-cell repertoire post allogenic bone marrow transplantation

The paper tracks T-cell repertoire changes post allogeneic bone marrow transplantation, finding that CDR3 sequences remain donor-derived while V-gene usage shifts toward the recipient over time.

Prediction of specific TCR-peptide binding from large dictionaries of TCR-peptide pairs

The paper introduces ERGO, a TCR-peptide binding predictor using NLP and deep learning, which outperforms existing methods in multiple benchmark tests.

Single haplotype admixture models using large scale HLA genotype frequencies to reproduce human admixture

This paper develops an HLA-based method to resolve population admixture using haplotype frequencies, identifying 8 ethnic groups that explain 90% of variance in 56 populations.