Proposal

Narrative

The growing volume of data generated in research coupled with the increasingly complex questions being addressed has given rise to new fields and professions like computational and data science which involve technologies like Deep Learning, harnessed by AI systems like ChatGPT and Google Lens. There's a need to equip our students with skills in computational modeling and AI-driven data analysis to be competitive for diverse job opportunities in academia and industry. Therefore, we aim to establish a multifunctional computational lab in the Jupiter Brain Institute and to enhance the existing computer labs in Boca by introducing high-end computers. This will facilitate: a) the teaching of existing (e.g. AI Applications in Biology IDS 4139) and new courses (e.g. Computational Neuroscience, Computational Biochemistry & Drug design to be developed by Drs. Pena and Fenollar Ferrer) simultaneously on both campuses b) provide hundreds of students working on Masters/PhD or Honors theses with a resource to accelerate their data analyses and generate professional figures for presentations & publications c) mitigate challenges associated with multiple campuses. Computer-based courses taught in a hybrid format combining in-person and remote instruction with multiple lab sections provide flexibility to students in choosing locations & times. Most undergraduate students (>500) majoring in Neuroscience and Behavior are located in Boca, but most research opportunities are in Jupiter. Offering more courses in Jupiter would allow Boca students to take them and pursue research in Jupiter without commuting between campuses on the same day; students residing north of Boca can take courses in Jupiter avoiding a southbound commute. In summary, the computer labs will be of value to hundreds of students from various departments with interests in Neuroscience, Bio-/Biomedical Sciences and generate new GTA opportunities on both campuses that align with Pillars of FAU’s strategic plan.

Facilities

Biological Sciences Computer labs in Boca (SC 177 and SC176) currently have each 17 workstations, equipped with standard computers that are not suitable for high-volume data processing. The third floor of the Brain Institute has designated space for 24 workstations (MC22/NR31) to develop a Program in Computational Brain Science and Health and serves as both a workspace and an educational facility. By acquiring suitable computers and connecting them to the Jupiter High-Performance Cluster via fiber optics, we can offer transferring and processing of high-volume data and employ hands-on courses in AI methods such as deep learning and simulation-based inference algorithms, which play a central role in estimating parameters from biological systems. The dedicated high-performance cluster in Jupiter, named KoKo, is well suited for these computations and fiber-optic ports are already available at the workstations in the Brain Institute. Additionally, there is a 10Gb/s connection between the Boca and Jupiter campuses at FAU. Via OIT-approved usage of remote desktop, this connection from Boca to Jupiter could also control of the newly installed computers from this proposal in Jupiter from Boca and provide a convenient link to KoKo for high throughput computations if necessary. We will re-distribute 12 computers from the Boca lab to Jupiter, and equally distribute 24 newly purchased high-performance computers between Boca and Jupiter to save costs and allow large data processing in both locations without remote or HPC cluster access as well via the HPC cluster in OIT-approved cases.

Hardware Requirements

24 high-performance Linux/Windows Precision 5860 Tower computers composed of a minimum of a 10-core processor of the order of 3.7GHz and NVIDIA CUDA Cores of the order of 6GB of GPU Memory which would allow 3 concurrently users in our estimations as well as serial and parallel computations both CPU- and GPU-based (12 in Jupiter with optic fiber cable cards to access to HPC cluster, 12 in Boca, without optic cable card but remotely accessing Jupiter through the 10Gb/s link) - allows for high volume data processing. 12 Mellanox Optical Fiber cards

Software Requirements

Neurosim 5, institutional license, 1-time upgrade purchase of our existing perpetual 40 licenses for unlimited licenses allowing to run the software on the 24 Jupiter and 37 Boca computers and could also be delivered through VDI-like server programs to PC and Mac users on or off campus. Course usage: Practical Cell Neuroscience course (PCB 4843C/ 6417C), Neurophysiology (PCB 4832C/6835C) Adobe Creative Cloud for professional photo & video processing and graphic design, 20 annual licenses of shared devices for 3 years. Course usage: Computer Graphics for Biologists (BSC 6455), Image and Video Processing and Vision in Marine Environment (OCE 5266), Scientific Communication (BSC 6846/4934) Matlab for programming & numeric computing, free via FAU license. Course usage: Artificial Intelligence Applications in Biology (IDS 4139), Experimental Design and Biometry (PCB 6456), in development courses: Computational Neuroscience, Microbial Community Analysis, Computational Biochemistry & Drug design R Studio for programming, free via FAU Apps. Course usage: Artificial Intelligence Applications in Biology (IDS 4139), Microbial Community Analysis (Course Prefix TBD), Experimental Design and Biometry (PCB 6456), Honors Experimental Design and Data Analysis (BSC 3452C), in development courses: Computational Neuroscience, Microbial Community Analysis, Computational Biochemistry & Drug design

Personnel Costs

none

Other Costs

none

Timeline

The goal is to set up the new computers with software on both campuses during the Spring/Summer 2024 semester. This will allow the offering of existing and newly developed computer courses (non-exhaustive list see below) as early as Fall 2024, with 2-4 lab sections in Boca (34- 68 student/semester) and 1-2 sections in Jupiter (24-48 students/per semester). However, the numbers of lab sections and the seat distribution between undergraduates and graduates can be easily adjusted on an as needed basis. Courses would be repeatedly offered in Fall or Spring semesters. Practical Cell Neuroscience, PCB 6417C & PCB 4843C, Instructor Godenschwege. AI Applications in Biology, IDS 4139, Instructor Pena Computer Graphics for Biologists, BSC 6455, Instructor Kajiura Image and Video Processing and Vision in Marine Environment, OCE 5266, Instructor Kajiura Experimental Design and Biometry, PCB 6456, Instructor McCoy Computational Neuroscience course, in development for undergraduates and graduates, Instructor Pena Microbial Community Analysis, in development for undergraduates and graduates, Instructor Fernandes Computational Biochemistry & Drug design, in development, Instructor Fenollar Ferrer

Sustainability

This computer set-up is designed to be sufficient for many years and have the capacity to run virtually any program, supporting the development of new courses that do not involve the above-listed software. We expect the enrollment to computer based courses to increase over the coming years, since these courses provide flexibility to the student, as well as the interest in having a hands-on instruction on how to perform programming or use AI technology is drastically increasing, ever since release publicly accessible apps allowing to identify objects, edit photos, or write comprehensive texts on virtually any topics like Google lens, PIXLR, ChatGPT, Segment Anything, Python tools for simulation-based inference, and PyTorch or TensorFlow for deep learning. The labs are also expected to be heavily used by undergraduate and graduate students who don’t have access to professional image processing to prepare images for publications and presentation or to computers capable of analyzing their data. There are annually over 200 students in the Biology Department working on their Masters, PhD or Honors thesis. The proper analysis of video and photographic data of sharks and endangered primates using AI and machine learning tools by Boca students, e.g., in the Kajiura, Porter and Detwiler labs is currently hampered by lack of access to suitable computers. Similarly, neuroscience, bio-medical and biochemistry students located in Jupiter from various departments have limited access to AI-ready computers to analyze for example experimental behavioral data from various model organisms or to perform computational biophysical chemistry used in drug design.

Resource Matching

Some of the required equipment is already provided by the Biology Department and the Brain Institute. Both Biology computer labs have combined 34 workstations with chairs, desks and standard computers. The Jupiter campus has 24 workstations with internet ports as well as optic fiber ports in addition to a significant UPS (uninterruptible power supply) upgrade that is already being prepared in the building but no computers. As a cost-saving measure and to allow for large data processing on both locations without and with the involvement of the HPC cluster (in case of time-consuming simulations), we will re-distribute 12 computers from the current Boca labs sufficient for some courses (e.g. Practical Cell Neuroscience, PCB 6417C & PCB 4843C) to Jupiter and equally distribute 24 newly purchased high-performance computers between Boca and Jupiter for courses that require high-performance computers (e.g. AI Applications in Biology, IDS 4139).

Implementing Organization

Biological Sciences Department in collaboration with the Jupiter Brain Institute.