Human Stories in AI: Tommy Tang

Overview

Tommy Tang details his journey from a low-income background in China to becoming the Director of Computational Biology at Imunitas Therapeutics, emphasizing the critical role of self-directed learning in computational biology. He highlights his experience with single-cell sequencing, machine learning (logistic regression, random forest), and workflow languages like Snakemake, applied to cancer immunology research and therapeutic development, including a CD161 antibody in Phase 1 clinical trials.

Key takeaways

Tommy Tang's career highlights the power of self-directed learning, transitioning from wet lab biology to computational roles through online courses and persistent effort.
Imunitas Therapeutics leverages single-cell genomics and machine learning (logistic regression, random forest) to identify cancer therapeutic targets, with a CD161 antibody (IM9) in Phase 1 trials.
The PD-1/PD-L1 pathway is a critical immune checkpoint where tumors can exploit PD-L1 to inhibit T cell activity, a mechanism targeted by immunotherapies.
Large-scale single-cell data analysis relies on cloud computing (Google Cloud) and workflow management tools like Snakemake due to data volume and complexity.
Accurate cell type annotation in single-cell data is challenging due to the continuous nature of cell states and the lack of definitive ground truth in complex biological systems.
Developing 'meta-skills'—the ability to learn new skills effectively—is crucial for navigating the evolving landscape of AI and computational biology, as demonstrated by Tommy Tang's approach to learning deep learning.

Chapters

0:00 Introduction to Human Stories in AI and Tommy Tang

Series features AI experts' career journeys.
Tommy Tang is Director of Computational Biology at Imunitas Therapeutics.
He has over 10 years of computational and 6 years of wet lab experience.
Focus on reproducible research and open science.

2:17 Early Life and Education in China

Born in a small town in South China to a low-income family.
Mother emphasized the importance of education for life change.
Ranked #1 in county high school and attended a top university in Shanghai.
Opportunity to learn about global opportunities and apply for US graduate schools.

5:00 Transition to Computational Biology at University of Florida

Arrived in the US on August 8, 2008, to pursue a PhD at the University of Florida.
Initial training was in wet molecular and cancer biology.
First computational challenge involved analyzing a TRIP sequencing dataset that crashed Excel.
Began self-teaching computational skills through online courses (edX, Coursera, Udacity).

6:53 Postdoctoral Research at MD Anderson Cancer Center

Moved to MD Anderson Cancer Center as a computational biology postdoc.
Worked with Dr. R.A. DePinho on the Cancer Genome Atlas (TCGA) project, analyzing large-scale genomic data.
Learned various sequencing data analyses (whole exome, whole genome, RNA-seq, ChIP-seq, bisulfite sequencing).
Developed proficiency in workflow languages like Snakemake for processing high-throughput ChIP-seq data.

13:48 Senior Bioinformatician and CID C Project Lead

Became a Senior Bioinformatician at Harvard FAS Informatics.
Analyzed single-cell RNA and ATAC sequencing data in collaboration with neuroscience labs.
Led the NIH-funded CIDC (Cancer Immunological Data Commons) project.
Processed clinical trial data from four cancer centers (MD Anderson, Dana-Farber, Mount Sinai, Stanford) on the cloud.

16:48 Director of Computational Biology at Imunitas Therapeutics

Joined Imunitas Therapeutics to establish computational biology capabilities.
Company focuses on using single-cell genomics to find new therapeutic targets for cancer.
Their lead program, IM9, is a CD161 antibody in Phase 1 clinical trials.
Identifies new targets for novel antibodies and supports biomarker analysis for clinical trials.

22:11 Machine Learning and Data Analysis Techniques

Utilizes 10x Genomics Chromium system for single-cell RNA and TCR sequencing.
Correlates gene expression profiles with T cell expansion and phenotype using logistic regression and random forest.
Explains PD-1/PD-L1 immune checkpoint mechanism and its role in T cell exhaustion.
Data processing and analysis are performed on Google Cloud using Snakemake pipelines.

Summary, takeaways, and chapters were generated by AI from the video's transcript and may contain errors. The video belongs to its creator, StatQuest with Josh Starmer.