Rajat Singhania

I was a key player in helping the company launch its hereditary diagnostics product covering panels from cancers to rare diseases. I built an NGS-based DNA mutation and CNV detection pipeline from upstream steps requiring raw data analyses to downstream steps, including variant calling. Also, I helped clinical scientists do variant filtering and interpretation. I was heavily involved in validating and QC of all the pipeline components and ensuring that all CAP and CLIA regulatory requirements were met. This project also involved building and validating bioinformatically a custom panel of genes covering the various disease conditions.

Various bioinformatics tools were used, including GATK, BWA, SAMtools, BEDtools, and IGV, along with the HGMD and Alamut databases. The primary programming language used was R, and bash scripting was done, too. The version control system Git was used. The project also involved communicating with various internal and external stakeholders.

This product impacted the care of hundreds of patients from referring physicians and hospitals. With the bioinformatics pipeline that formed the essence of this product, actionable variants and the associated disease were identified for personalized care.

During my postdoc at the prestigious Princess Margaret Cancer Center, I was a co-first author of a Nature paper that showed the utility of the novel cfMeDIP-seq assay developed by Professor Daniel De Carvalho in detecting and classifying cancer from a patient's cell-free DNA that can be obtained via a simple blood draw. I showed the proof of principle of this liquid biopsy approach by:

• Developing the bioinformatics pipeline that processed the raw NGS data from samples run with this assay.
• Creating a machine learning approach by conducting cross-validation and choosing the best technique among a generalized linear model (GLM), support vector machine (SVM), and random forest that showed the high sensitivity and specificity of the assay.

A variety of bioinformatics tools were used for the execution of this project, including BEDtools, Bowtie 2, IGV, MEDIPS, and DESeq2. The main programming language used was R, and bash scripting was done. Data analyses included dimension reduction techniques such as PCA and t-SNE. A Sun Grid Engine HPC cluster for PBS job scheduling was used.

This Nature paper formed the basis for the launch of the startup Adela where I made further contributions that earned me the CEO Award in 2022.