Research covers various topics, including cardiology and the study of heart disease. In addition, they collaborate with a prestigious research group in Murcia and Arrixaca. This group has numerous samples of heart tissue and blood.

However, the group needs funding to analyze these samples. In this context, they face two main challenges:

• First, get money to analyze the samples at the multiomic and single cell level.
• Second, manage and process the large amount of data obtained efficiently.

The research addresses areas where explaining algorithms and data is essential, as understanding the information collected helps solve problems.

In these fields, "black boxes" are not suitable, as they cannot answer such questions.

Although humans are adept at perceptual tasks, we find it difficult to find hidden answers in huge data sets and detect underlying patterns.

In their research, they need algorithms that scale horizontally, especially in machine learning interpretable. When facing problems with big data, there was limited storage and infrastructure to scale and run algorithms quickly.

Therefore, they prioritized having many CPUs.

In addition, they created explainable artificial intelligence algorithms, focusing on deep learning and generative models. Here, it is crucial to have advanced GPUs based on the Ampere architecture.

As they also design spatiotemporal simulation models, it is imperative that GPUs handle double precision computations efficiently.

Thus, the needs of the group were mainly three:

– Many CPU nodes to do parallel computing and apply their models of machine learning interpretable.

– Next-generation, double-precision GPUs for explainable AI models and simulation.

– A lot of storage, especially for the biomedical application that will be explained below and that allows for a significant amount of funding to be obtained through a European research project that involves researchers from all over the world.

Leaders in biomedical research

La Universidad Rey Juan CarlosBeing a young and dynamic institution, it strives to contribute to the vanguard of knowledge, an achievement that is reflected in national and international rankings, making this organization one of the leaders in biomedical research.

Currently, it has approximately 46.000 students spread over 5 campuses with 28 departments in multiple fields of arts, sciences, and literature.

The project "Machine Learning on Single Cell Multiomic to Relate COVID-19, Aging and Cardiopathies” was funded as part of the Union's response to the COVID-19 pandemic.

It involved the collaboration of centers such as CNB-CSIC, CEMBIO, Science Park of Madrid y MIT.

The research groups of the Universidad Rey Juan Carlos developed a preclinical animal model to study the effects of the cytokine storm. This severe immune reaction caused mortality in COVID-19 patients, especially the elderly and people with heart problems.

With this preclinical model, a platform was created to detect targets and design useful treatments not only for COVID-19, but also for other pathologies or future pandemics. The project allowed the creation of a supercomputing center to design and apply the necessary algorithms in their studies.

Leaders in biomedical research

An integration team was contacted to design a cluster with the necessary technical capabilities. Thanks to SIE's knowledge and experience in HPC, the ideal computer center for the investigations was configured, integrating machines acquired from Gigabyte.

Leaders in biomedical research

Knowledge transfer is carried out through the spin-off EBTC, Dilemma Lab SL This company was created in collaboration with a researcher from the Miguel Hernández University of Elche.

1.- Cellular aging. Revealing the mechanisms that produce it.

2.- Partial cell reprogramming. Studying the reversal of cell deterioration to improve quality of life in advanced stages.

3.- Heart disease. Using the cardiac signals and the techniques of the previous lines for its prevention and treatment.

4.- Oncology. Going deeper into its mechanisms for personalized treatment, studying the feasibility of applying the second line of research mentioned above.

Connectivity and storage

The A100 cards are interconnected via NVLINK buses, reaching speeds of 600 Gb/s. NVIDIA/Mellanox Infiniband HDR is used to connect equipment, allowing:

1. Interconnect 32 A100 cards through NVSwitch and Infiniband protocol.
2. Add 256 CPU cores (512 threads) with AMD Milan 7763 processors.
3. Access 576 TB of shared storage with high management speed. Access to this shared storage capacity of the cluster is done by native InfiniBand, with high management speed, through a data and metadata repository, to which access by BeeGFS, allows a great concurrent speed of the nodes and user management

The system can be expanded up to 1 PB in the future.

SIE mounts the system in the ecosystem HPC Ladon OS 8, based on Open Source pieces. Avoid paying for proprietary software and reduce costs. Fundamental tools include:

• RockyLinux 8.7: Base operating system, stable and updatable. It isolates the internal network from the external one using IP table.
• SLURM: Queue or workflow system used by leading clusters such as Mare Nostrum or Hyperion.
• Docker: Container system that allows individualizing applications without virtualizing the entire machine.
• CheckMK: Management console based on Nagios. It works through IPMI and monitors devices through SNMP.
• EasyBuild: Software creation and installation framework for managing scientific software on HPC systems efficiently.