AI and Statistics in Proteomics and Systems Biology – Interview with Professor Olga Vitek Ph.D.

Professor Olga Vitek has a deep understanding of statistics, machine learning, and computational biology. She puts her know-how to work to develop computational tools enabling high-quality proteomic analysis and systems biology approaches. She hopes to apply these tools to the quantitative analysis of large-scale mass spectrometry-based investigations and thereby advance our understanding of organismal function. In this episode, Olga and Parag discuss:

• Why statistics is important for experimental design
• How statistics and AI can help researchers understand biology
• Gaps keeping us from using AI and statistics to their maximum potential in biology

Find this episode on YouTube, Apple Podcasts, and Spotify.

Chapters

00:00 – 01:26 – Intro
01:27 –  04:26 – Why did Olga decide to apply statistics to biology and proteomics in particular?
04:27 – 06:13 – Factors leading to the adoption of statistics in proteomics
06:14 – 10:06 – Why do we need statistics for experimental design?
10:07 – 14:58 – How does statistics deal with observational experiments?
14:59 – 19:04 – Statistical principles Olga wishes more researchers were aware of
19:05 – 27:21 – How do we balance the use of AI models with the need for rigor and interpretability in our analyses?
27:22 – 36:11 – Combining data from multiple sources using tools that reason on the language of biological molecules
36:12 – 43:34 – In Olga’s dream future, how will researchers be using AI, statistics, and machine learning?
43:35 – 45:25 – What gaps are keeping us from achieving Olga’s dream?
45:25 – End – Outro

Resources

Statistical methods for studies of biomolecular systems website
Olga’s personal lab website.

Beyond protein lists: AI-assisted interpretation of proteomic investigations in the context of evolving scientific knowledge
Gyori and Vitek, 2024 discuss how AI can be used to interpret proteomics data and its biological meaning.

A Bayesian Active Learning Experimental Design for Inferring Signaling Networks
Ness et al., 2018 show how statistical methods can guide the selection of experiments that optimally enhance understanding.