Oversaw development of internal data analysis toolkits and platforms. Established comprehensive database of cis-regulatory elements and their cell/tissue specificities. Identified epigenetic signatures of chemotherapeutic agents in cell lines and bulk tissues (chromatin & dna methylation). Integrative multi-omic prognostic analysis of single-cell and bulk sequence oncology data. Development of libraries for single-cell coexpression networks and recptor-ligand interaction estimation. Statistical power analyses for observational clinical trials & prospective genetic studies. Estimation of diagnostic yield for potential clinical genetic products on the basis of known mutations, mutation rates, and known causal genes.

Applied unsupervised learning and supervised learning to high-throughput flow-cytometry data (>250,000,000 cells) to identify shifts in rare T-cell population between refractory and non-refractory tumors. Developed apps and workflows on the DNANexus platform for assembly and annotation of bacterial genomes, including crispr/cas system classification, AMR resistance gene identification, and locating bacteriophage insertions. Established novel methods for measuring the degree of response to treatment, using gene expression as a phenotype. Designed and implemented relational database for data & metadata related to bulk tissue and single-cell gene expression studies which enforces an ontological controlled vocabulary; implemented tools for validating correctness & importing curator-provided metadata. Leveraged gene and protein expression across several tumor cohorts to prioritize potential novel therapeutic targets.

As part of the Dan Geschwind lab at UCLA, I study the neurogenetics of Autism and ASD. In particular I develop methods to identify molecular endophenotypes by integrating gene and micro/lncRNA expression with epigenetic information (histone modifications, 3d structure, TF binding) and genetic mutations. My specific aim is to identify molecular endophenotypes underpinning ASD, and the convergent neurodevelopmental pathways which are implicated by these disruptions. An example of our work: https://www.biorxiv.org/content/10.1101/2020.03.05.965749v1

Bioinformatics: Developed software tools to evaluate the performance of RNA-seq aligners and assemblers on de novo gene fusion events, both in silico and with real cancer data. Created software for the estimation of low-level DNA mixtures (contamination, tumor, or fetal) which combines information from polymorphic sites with epigenetic information (through treatment with a restriction enzyme). Developed algorithm to detect large structural variation in admixed cfDNA from low input quantities (6ng). Medical informatics: Engineered a python machine learning framework for EMRs (sqlalchemy + pandas + scikit-learn) and applied it to predict adverse medical events which greatly extend inpatient duration

One of five developers responsible for private-cloud-based machine learning pipelines (discovery/automated validation and deployment/sample classification) for early diagnosis of ASD from whole-blood RNA-seq. Experience: Evaluation of feature extraction and classification pipelines (biomarker discovery), developing frameworks for efficient deployment of ML tests to AWS, use of cryptographic libraries for safe cloud storage.

Delivered informatics pipelines and spearheaded development of analytical tools for the discovery, analysis, and validation of disease-linked variation in humans from NGS data. Eliminated several analytic bottlenecks by proposing a more efficient, powerful, and general statistical method for rare variant association, and developing the software implementing it. Novel methods for identifying shared genetic architectures between traits (pleiotropy), adjusting heritability estimates from genotype data for linkage, and bounding the frequency/effect distribution for rare variants. GATK Developer.

• Improved current methods of assessing & calibrating the quality of analytic output, enabling progress by making it easy to identify positive and negative changes• Improved nonparametric statistical bias calculation, reducing false discoveries by 78% • Developed novel tools and improved existing tools to meet research demands, resulting in cleaner results and quicker turnaround times• Suggested improvements to core algorithms that lowered complexity of development

• Designed, developed, and piloted a non-parametric statistical approach for the prediction of clinical trial enrollment periods based on performance of known trials in database (TrialPredict)• In a continuing consulting capacity, oversaw implementation of pilot algorithm into a production database setting