Graph AI is a revolutionary technology that facilitates highly accurate learning by utilizing knowledge graphs within data sets, including those containing defects—a common occurrence in biotechnology. This technology is broadly applicable across a wide range of fields. 

Link:
Graph-based Relational Learning: NEC Technical Journal | NEC

In designing personalized vaccines, we aim to predict the likelihood of an immune response based on an individual's MHC alleles. We focus on developing models to predict MHC class II peptide binding and presentation, using self-supervised pretraining on large protein sequence datasets. Our work improves prediction accuracy and has potential applications in T-cell-based immunotherapy, offering new directions for training models in peptide binding and presentation.

Link:
BERTMHC: improved MHC–peptide class II interaction prediction with transformer and multiple instance learning | OP-CBIO210426 4172..4179 (neclab.eu)

The complex characteristics of cancer tissue, particularly tumor genetic intra-heterogeneity, combined with sequencing and alignment artifacts make somatic variant calling a challenging task. To address this, we developed the NeoMutate framework, which integrates seven supervised machine learning algorithms to leverage the strengths of multiple variant callers based on a unique set of biological and sequence features.

Link:
NeoMutate: an ensemble machine learning framework for the prediction of somatic mutations in cancer | BMC Medical Genomics | Full Text (biomedcentral.com)

We developed a digital twin simulation using a database of HLA haplotypes from approximately 22,000 individuals to model the effectiveness of different epitope hotspots in a diverse population. This method identifies the optimal set of hotspots for maximum global coverage. By integrating the NEC Immune Profiler's predictions on antigen presentation, cell surface binding, and immunogenicity with a robust Monte Carlo and digital twin simulation, we pinpoint a subset of epitope hotspots for potential use in broadly protective vaccine formulations.

Link:
Artificial intelligence predicts the immunogenic landscape of SARS-CoV-2 leading to universal blueprints for vaccine designs - PMC (nih.gov)