Periodic Reporting for period 1 - BioExcel-3 (BioExcel Centre of Excellence for Computational Biomolecular Research)
Período documentado: 2023-01-01 hasta 2023-12-31
Computing is a critical resource for high-impact research and commercial applications, in particular in Life Sciences. Exascale computer systems enable ever faster modelling and simulations, with paramount influence on health/medical applications, drug development, efficient drug delivery, biotechnology, environment, agriculture, food industry, and not least education.
BioExcel CoE was established already in 2015. In the current phase three the centre contitnues to:
* Support & maintain key European open-source applications that have been selected by researchers as their tools of choice (measured by applications and scientific citations, with periodic re-assessment), and improving their performance, reliability, and quality to meet academic and industrial needs.
* Refine the user-driven development where all efforts will be prioritised based on scientific impact and ensuring European researchers will be able to deploy on EuroHPC resources efficiently.
* Educate and enable users how to exploit modern technologies, including ensemble algorithms, AI approaches, and convergence of high-performance computing, high-throughput computing, and high-performance data analytics as well as providing the workflows necessary to achieve this.
* Develop and delivering state-of-the-art online and hands-on tutorials, documentation, and hands-on training resources for next-generation workforce development, in particular PhD students.
* Create a new BioExcel Ambassador Program to enable the greater community to more explicitly channel input to code developers and help disseminate knowledge and training.
BioExcel CoE was established already in 2015. In the current phase three the centre contitnues to:
* Support & maintain key European open-source applications that have been selected by researchers as their tools of choice (measured by applications and scientific citations, with periodic re-assessment), and improving their performance, reliability, and quality to meet academic and industrial needs.
* Refine the user-driven development where all efforts will be prioritised based on scientific impact and ensuring European researchers will be able to deploy on EuroHPC resources efficiently.
* Educate and enable users how to exploit modern technologies, including ensemble algorithms, AI approaches, and convergence of high-performance computing, high-throughput computing, and high-performance data analytics as well as providing the workflows necessary to achieve this.
* Develop and delivering state-of-the-art online and hands-on tutorials, documentation, and hands-on training resources for next-generation workforce development, in particular PhD students.
* Create a new BioExcel Ambassador Program to enable the greater community to more explicitly channel input to code developers and help disseminate knowledge and training.
The newest pre-exascale machines have come online and the codes are ready to run on those. In particular, there have been significant efforts to make GROMACS perform well on LUMI. This has been a challenge, not so much because of the new architecture of the AMD GPUs, but rather because of the software stack, which has had some stability issues. By now, GROMACS achieves good performance on LUMI and we have scaling benchmarks to back this up. Notably, while BioExcel as a whole was designed for track 2 of the CoE mechanism since most users have a strong need for ensemble parallelism, GROMACS in particular achieves outstanding scalability, performance and portability to all major GPU hardware. Scaling benchmarks have been run for both GROMACS and HADDOCK on different HPC machines, including LUMI. They show scaling up to 128 GPUs or 2048 CPU cores. In BioExcel-3 we have moved to systematic user driven development. We have sent out and analysed questionnaires with similar structure for all the codes. Work on new features is prioritised based on the outcomes.
Usability and reproducibility are being improved in HPC supercomputers, with the work on a new set of BioBB and PyCOMPSs Spack packages, easing the installation of our workflows in the new EuroHPC machines. The combination of AI-based workflows with EuroHPC supercomputers and HPC-focused workflow managers will enable real science with automation and efficient use of resources, reducing the time-to-result, similarly to what was achieved in BioExcel-2 with our Pre-exascale HPC approaches for molecular dynamics simulations on the Covid-19 research.
By conducting user surveys and analysing the results of those, we have demonstrated our commitment of understanding and integrating user needs into the software development process. The surveys collected for the core BioExcel codes have helped shape the roadmap and training plans, reflecting a user-driven approach to software enhancement. The newly started Ambassador Program and community-driven activities play a strategic role in engaging the user community and understanding and addressing specific requirements of different user groups. In the first year of the project, we successfully built an ambassador community of 21 country representatives. This program has already successfully expanded our community engagement through workshops and activities. A first outcome of this programme, the “Carpathian edition” was a GROMACS/HADDOCK training workshop organised in Bratislava in October 2023, organized jointly with NCCs of Slovakia, Czech Republic, Austria and Hungary. Our Ambassador Program has already facilitated direct interactions with users with more activities already planned.
The 8th edition of our Flagship training event “BioExcel Summer school” was organised in Sardinia in September 2023 for 30 students containing lectures and hands-on sessions of the core tools. The three best posters were selected by vote and their authors were given an opportunity to present their research in an additional BioExcel webinar in late 2023. The Summer school average of feedback from participants was excellent 9,5/10. Two other on-site training events were organised: A fully on-site event, the Spring School on Computational Chemistry, was organised in Finland already in April. In October, a hybrid meeting for the new ambassadors linked with a training workshop, was organised in Bratislava, together with the BioExcel all hands meeting. The successful BioExcel webinar series was relaunched in April 2023. A total of eight new webinars were organised during the reporting period, see https://bioexcel.eu/category/webinar/. Additionally, BioExcel partners participated in numerous other events promoting the core software and by direct support via the Ask BioExcel forum.
A high level of dissemination of BioExcel events, news, job postings and related activities has happened mainly through the three key social media platforms (X/Twitter, LinkedIn, YouTube) as well as through the bi-monthly BioExcel newsletter. This is reflected in the increased numbers of users, followers and subscribers for each of the channels. BioExcel has over 3,500 followers on X/Twitter (7% increase in 2023), over 2,100 followers on LinkedIn (45% increase in 2023) and over 3,700 subscribers of the YouTube channel (26% increase in 2023) with more than 191,000 total views of overall content. In addition, the BioExcel mailing list which is used to distribute the bi-monthly newsletter had approximately 2,800 subscribers at the end of 2023 (19% increase).
Usability and reproducibility are being improved in HPC supercomputers, with the work on a new set of BioBB and PyCOMPSs Spack packages, easing the installation of our workflows in the new EuroHPC machines. The combination of AI-based workflows with EuroHPC supercomputers and HPC-focused workflow managers will enable real science with automation and efficient use of resources, reducing the time-to-result, similarly to what was achieved in BioExcel-2 with our Pre-exascale HPC approaches for molecular dynamics simulations on the Covid-19 research.
By conducting user surveys and analysing the results of those, we have demonstrated our commitment of understanding and integrating user needs into the software development process. The surveys collected for the core BioExcel codes have helped shape the roadmap and training plans, reflecting a user-driven approach to software enhancement. The newly started Ambassador Program and community-driven activities play a strategic role in engaging the user community and understanding and addressing specific requirements of different user groups. In the first year of the project, we successfully built an ambassador community of 21 country representatives. This program has already successfully expanded our community engagement through workshops and activities. A first outcome of this programme, the “Carpathian edition” was a GROMACS/HADDOCK training workshop organised in Bratislava in October 2023, organized jointly with NCCs of Slovakia, Czech Republic, Austria and Hungary. Our Ambassador Program has already facilitated direct interactions with users with more activities already planned.
The 8th edition of our Flagship training event “BioExcel Summer school” was organised in Sardinia in September 2023 for 30 students containing lectures and hands-on sessions of the core tools. The three best posters were selected by vote and their authors were given an opportunity to present their research in an additional BioExcel webinar in late 2023. The Summer school average of feedback from participants was excellent 9,5/10. Two other on-site training events were organised: A fully on-site event, the Spring School on Computational Chemistry, was organised in Finland already in April. In October, a hybrid meeting for the new ambassadors linked with a training workshop, was organised in Bratislava, together with the BioExcel all hands meeting. The successful BioExcel webinar series was relaunched in April 2023. A total of eight new webinars were organised during the reporting period, see https://bioexcel.eu/category/webinar/. Additionally, BioExcel partners participated in numerous other events promoting the core software and by direct support via the Ask BioExcel forum.
A high level of dissemination of BioExcel events, news, job postings and related activities has happened mainly through the three key social media platforms (X/Twitter, LinkedIn, YouTube) as well as through the bi-monthly BioExcel newsletter. This is reflected in the increased numbers of users, followers and subscribers for each of the channels. BioExcel has over 3,500 followers on X/Twitter (7% increase in 2023), over 2,100 followers on LinkedIn (45% increase in 2023) and over 3,700 subscribers of the YouTube channel (26% increase in 2023) with more than 191,000 total views of overall content. In addition, the BioExcel mailing list which is used to distribute the bi-monthly newsletter had approximately 2,800 subscribers at the end of 2023 (19% increase).
BioExcel core applications continue to be at the forefront of HPC computing with support for all existing major hardware systems. Co-design activities ensured portability, scalability and extreme performance. Workflow solutions were successfully used for pre-exascale productions runs and see wider adoption. All of our codes as well as main workflow platforms are included in the EU Innovation Radar. Usage of our tools continues to grow as shown by the increasing number of publications referencing the codes. BioExcel is more and more recognized as a provider of highly sought after expertise not only in Europe but also US, Japan, India, South-East Asia. Our training events are regularly oversubscribed multi-fold. The training program continues to develop with very effective onsite and remote capabilities. Training material has grown substantially.