Chemistry and Materials Science Seminar: A Journey of AI in Chemistry and the SPECTRUM Project

Event Image
scms seminar banner

Chemistry and Materials Science Seminar
From Foundations to Frontiers: A Journey of AI in Chemistry and the SPECTRUM Project

Dr. Christopher Collison
Professor

School of Chemistry and Materials Science, RIT

Register Here for Zoom Link

Abstract:
The intersection of artificial intelligence (AI) and chemistry is at a nascent stage, with much of the AI realm engrossed in the relentless pursuit of novel models and transformative architectures. Despite these rapid advancements, the practical integration of AI into everyday chemistry remains an area ripe for exploration. The application of AI in predicting chemical properties comes with an inherent question: how trustworthy are these predictions? Herein lies the merit of “explainable AI”. It not only offers predictions but also, critically, provides reasons for them, which might lead to the trust of chemists. This would ensure that AI models align with chemical intuition and principles. My sabbatical journey began by adapting explainable AI methodologies, initially honed for molecular solubilities, to the realm of optical spectroscopy. The expected observation surfaced that molecular scaffolds, rather than subtle modifications to peripheral functional groups, predominantly influence absorbance properties. This observation leads to a complication and a revised approach to explainable AI for photoactive materials; a universal approach to AI model explanation may often fall short when addressing the nuances of diverse molecular systems and their varied applications. Inspired by these findings, our focus transitioned to the domain of “inverse design”, culminating in the creation of the SPECTRUM methodology: "Substructure-Preserving Emphasized Counterfactuals for Topological Refinement and Unraveling Molecular properties". With a fixed molecular scaffold set in place, our approach would now probe how peripheral side-groups might be changed to accentuate one property, without compromising another crucial optical attribute. In this seminar, attendees will learn the foundations of AI and its synergy with chemistry. I'll broadly demystify the AI algorithms, share invaluable insights garnered from our deeper-dive into photoactive molecules, and cast a gaze towards the promising horizons of this interdisciplinary union.

Intended Audience:
Undergraduates, graduates, experts. Those with interest in the topic.

To request an interpreter, please visit myaccess.rit.edu


Contact
Nathan Eddingsaas
Event Snapshot
When and Where
October 24, 2023
12:30 pm - 1:45 pm
Room/Location: 2305
Who

This is an RIT Only Event

Interpreter Requested?

No

Topics
faculty
research