| Yoram Louzoun’s Lab | Department of Mathematics | Bar Ilan University |


The lab has five main groups, all dealing with different aspects of machine learning, graph analysis, computational immunology, and stochastic processes. We also study some specific projects beyond these groups. The groups are:

  1. B and T lymphocyte receptor repertoire analysis
  2. Bone marrow transplants.
  3. Microbiome analysis
  4. Network-based machine learning
  5. Stochastic processes.

Other subjects, we study are epidemiological models and epitope binding predictions.

Lymphocyte Receptor Repertoire

Maintenance of long-term memory in the adaptive repertoire is a balance between completeness of coverage, the need to maintain large clones against common pathogens and parsimony of memory storage in a finite number of memory cells. The B and T cells, key players in the adaptive immune system. These receptors are diversified through a sequence of mechanisms that maximize this diversity to enable a potential response to every presented peptide. Until recently, immune receptor repertoire studies were limited by low throughput of classic sequencing. Now, with the rapid advance of Next-Generation sequencing, we are able to perform large scale BCR and TCR repertoire studies, with a high throughput of over a million sequences per host. However, novel mathematical and machine learning methods are required to analyze such data.

In collaboration, with multiple labs in Israel and the US, we develop mathematical and computational tools to analyze large scale repertoires, including:

  1. Prediction of the receptors that can bind a target
  2. Use repertoires as diagnostic tools
  3. Correlate T cell receptors with MHC
  4. Study the stochastic dynamics of repertoire generation
  5. Develop optimal methods to represent these repertoires.

Bone marrow transplants

In close collaboration with Ezer-Mitzion in Israel and be the match in the US, we study methods to improve HLA allele imputation and donor-patient matching. The work led to a novel imputation and matching algorithm (GRIMM) now tested on large donor registries. We also develop algorithms to predict the usefulness of new recruits to registries and methods to handle registries with mixed ethnicities.

Microbiome analysis

We perform multiple microbiome computational projects in collaborations with the Koren lab in Tzfat, and other labs. We develop diagnostic and prognostic tools based on a combination of microbiome and machine learning, as well as advanced methods to predict microbiome dynamics. These projects combine advanced machine learning with the complex dynamics of the microbiome.

Graphs and machine learning

We study the interplay between the structure of networks and the content of the nodes. We study many issues, including:

  1. Analysis of the influence of nodes’ content on the dynamic of the structure of the network, and the emergence of the network’s properties.
  2. Graph Convolutional Networks to predict the future dynamics of networks and the content of nodes
  3. Detecting groups in networks and their evolution.
  4. Predicting decisions based on network topology

The analysis is based on a combination of theoretical tools, advanced machine learning and large scale measurements in observed networks. It is performed in collaboration with multiple companies.

Stochastic Processes and Population Dynamics

Birth death processes can have a very non-intuitive behavior. We study the stochastic population dynamics of metapopulations. This analysis is mainly focused on the human HLA locus. We study the population genetics of human sub-population and the selection forces affecting HLA alleles and haplotypes. This is performed in collaboration with the NMDP. We develop theoretical models as well as simulations to understand the observed human population composition.