Andrew Prescot

Intro

Data Science Skills:

• Python programming fundamentals with specific emphasis on the Python data stack (NumPy, Pandas, Scikit-learn) as well as visualization libraries and tools (Matplotlib, Seaborn)
• SQL (postgreSQL) querying, aggregating, grouping, and joining data
• Experimental design and A/B testing
• Model preparation and exploratory data analysis including data cleaning, exploration, and feature engineering
• Supervised machine learning using Scikit-learn with emphasis on classification, regression, similarity, random forest, support vector machines, boosting models
• Unsupervised machine learning using Scikit-learn focusing on clustering using k-means, hierarchical, DBSCAN, and Gaussian mixture models. Performance evaluation of clustering models
• Dimensionality reduction via PCA, t-SNE, and UMAP
• Deep learning, neural network architectures, supervised and unsupervised neural network models
• Implementation and hyperparameter tuning of convolutional neural networks for computer vision (Keras, Tensorflow)
• Hands-on practical experience installing and working with GPUs to develop models

Work

A comparative study of atmospheric nitrogen dioxide (NO₂) concentration in Salt Lake City and Honolulu. This project aimed to evaluate temporal NO₂ changes between 2009-2019 as well as seasonal effects. Similar to altitude-of-residence, elevated atmospheric NO₂ concentration has been identified as a risk factor for increased odds ratio for suicide.
Salt Lake's Dirty Air

Evaluation of supervised machine learning models for epitope prediction used in COVID-19 vaccine development.
COVID-19/SARS B-Cell Epitope Prediction

Assessment of dimensionality reduction techniques and unsupervised machine learning models for clustering of human electroencephalogram (EGG) signals recorded from five distinct seizure and non-seizure conditions.
Unravelling EEG Signals via Unsupervised Machine Learning Approaches

The aim of this project was to implement and tune a convolutional neural network (CNN) model that enables automatic detection of malignant tissue within histopathological scans of human lymph node sections (PatchCamelyon dataset). This type of CNN model would be of enormous help to medical researchers and practitioners, allowing for binary classification on a scale that will surpass human capability with the added benefit of removing error due to human factors.
Implementation and Evaluation of Artificial Neural Network Models for Automatic Detection of Malignant Tissues within Histopathological Scans of Lymph Node Sections

Other Work

Over the years I have had a keen interest in the role of non-invasive medical imaging modalities to investigate the pharmacodynamic effects of novel agents to treat psychiatric illness and neurological disease. Specifically, I have used magnetic resonance spectroscopy (MRS) techniques to perform these types of investigations, which essentially uses clinical MRI machines to interrogate tissue chemistry rather than acquire the more familiar brain (or any other tissue) grayscale images. For the brain, MRS can measure import amino acid neurotransmitter compounds including GABA and glutamate, which control the overall cerebral inhibitory-excitatory balance, respectively. If you are interested, click on the following link to see published examples of MRS studies that we performed to evaluate the response of brain GABA to CPP-115, a next-generation GABA-aminotransferase (AT) inhibitor that was developed to treat seizure disorders. Compared to other widely-used GABA-AT inhibitors such as vigabatrin, CPP-115 can be administered at much lower doses, leading to a much more favorable toxicity profile especially regarding ocular damage.

Technologies taken from many different domains will support and expedite the drug discovery process!

Click the link below to see two selected first author publications published in the Nature family of journals:

About

Medical Imaging Ph.D (Biophysics) and Data Science enthusiast possessing a unique educational and professional background encompassing organic chemistry, medical imaging research/development, and the practical application of data science and machine learning (ML) methodologies for solving biological problems. Especially interested in the emerging critical role of artificial intelligence and ML technologies for enhancing biomedical research and personalized healthcare.

For downtime I enjoy playing the guitar, watching Manchester United, golfing, skiing, continually failing to create authentic Indian food, and hiking with our amazing 2 year-old Golden retriever, Cleo

Contact

andrewprescot@gmail.com

• fa-linkedin
• GitHub

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';