The Adam Widman Lab

The Widman lab within the Digital Oncology Program at MSKCC is using machine learning and genomics to engineer the next-generation of ultrasensitive, highly scalable liquid biopsies for early cancer detection, treatment monitoring, and detection of minimal residual disease (MRD) after surgery for early-stage cancer.

Liquid biopsy via circulating tumor DNA (ctDNA) offers the potential to serve as a critical biomarker in many settings within solid tumor oncology, such as the early detection of cancer, monitoring of treatment response (advanced disease, neoadjuvant, and adjuvant therapy settings), and detection of MRD after surgery. However, in many early-stage cancers and low volume treatment settings, ctDNA signal is sparsely diluted and thus difficult to detect among cell-free DNA (cfDNA) from noncancerous cells. Current ctDNA detection practices use deep targeted sequencing of a few informative sites such as patient-specific bespoke panels. These platforms are limited, as many commercial panels lack the sensitivity to reliably detect early-stage cancer that can be seen on imaging, much less low volume disease states like MRD. Further, the matched tissue requirement of many panels serves as a significant barrier to real world sensitivity, because patients often do not have matched tumor tissue available for analysis. Liquid biopsies are therefore not routinely used in many settings where they could be applicable, even at tertiary academic centers. For example, there are no FDA-approved liquid biopsy tests to monitor for MRD in solid tumors.

To overcome barriers to widespread use, we reasoned that the next generation of liquid biopsy tests would need to be 1) ultrasensitive, to detect and quantify tumor fraction (TF) in low-volume disease settings, and ii) tumor-naïve and highly scalable, to be deployed rapidly in all treatment settings where ultrasensitive liquid biopsy is needed. 

Our approach to liquid biopsy leverages advanced computational algorithms to extract sparse ctDNA signal from plasma whole genome sequencing (WGS). Plasma WGS is rapidly declining in cost and is highly scalable, as it can be run in virtually any genomics lab. WGS captures the breadth of cancer mutations (termed single nucleotide variants [SNVs]) and cancer aneuploidy (copy number variants [CNVs]), two important hallmarks of cancer that distinguish ctDNA from benign cfDNA. Our lab leverages novel deep learning strategies for detecting cancer mutations as well as data science and statistics methods to quantify anueploidy in cfDNA through a myriad of features including read-depth, B-allele frequency, and fragmentomics.

We are looking for highly motivated individuals with experience in machine learning and / or bioinformatics who want to make a sweeping impact on cancer detection and treatment monitoring. We are recruiting for technician, analyst, and software engineering roles.