Artificial Intelligence in MS Instrumentation and Applications

This Tuesday afternoon oral session, on the subject of artificial intelligence (AI) in MS instrumentation, will be held from 2:30 pm to 2:50 pm in Ballroom B. Jesse G. Meyer of Cedars-Sinai Medical Center will chair and preside over this afternoon session, which includes six talks addressing a variety of approaches and applications of AI in MS.

At 2:30 pm, Qing Zhong of ProCan, Children’s Medical Research Institute of The University of Sydney in Australia, will begin the session with a talk on pan-cancer proteomic mapping of 949 human cell lines to reveal principles of cancer vulnerabilities in cells. Raw proteomic data were acquired by data independent acquisition mass spectrometry (DIA-MS) with data being analyzed using a deep learning-based algorithm (DeeProM).

Next at 2:50 pm, Sanjay Iyer of Purdue University in West Lafayette, Indiana, will discuss a machine learning liquid chromatography-tandem mass spectrometry (HPLC–MS/MS) platform based on diagnostic ion-molecule reactions for structural identification of unknown compounds. An automated machine-learning (ML) guided HPLC–MS/MS system was demonstrated in this research to predict the most probable functionalities in the unknown analytes of samples.

William E. Fondrie from Talus Bioscience in Seattle, Washington, will then give a talk at 3:10 pm on “The Need for Speed,” discussing compressed sensing with generative models enabling ultrahigh-throughput (HT) mass spectrometry. Fondrie reports on a data-independent acquisition (DIA) mass spectrometry method using deep learning that was demonstrated to acquire high throughput proteomic data at a rate of 5 minutes per sample.

The topic of MetFID or using convolutional neural network (CNN)–based compound fingerprint prediction for metabolite annotation will be addressed next at 3:30 pm, by Shijinqiu Gao of Georgetown University in Washington, DC. This talk presents a new neural network–based tool that predicts compound fingerprints based on MS/MS data to identify metabolites.

In the fifth talk, at 3:50 pm, Tingting Zhao of the University of British Columbia in Vancouver, British Columbia, will discuss library-free cleaning of chimeric MS/MS spectra for metabolomics. This work shows the development of an MS/MS purification workflow (DNMS2Purifier) that automatically identifies and removes contamination fragments from experimental MS/MS spectra.

Md Inzamam Ul Haque of The University of Tennessee in Knoxville, Tennessee, will close out the session at 4:10 pm with a talk describing the prediction of prostate cancer directly from tissue images using deep learning on mass spectrometry imaging and whole slide imaging data. Here, visual imaging features obtained through deep learning from whole slide-stained tissue are combined with imaging MS (IMS) data to correlate the imaged chemical signature within tissue samples.

Comments are closed.