Research Projects

REPRODUCIBILITY OF SAGA ALGORITHMS

2021 - 2022

Probabilistic Modeling; Machine Learning; Reproducibility; Epigenomics; Genomics

Research on evaluating the reproducibility of predictions from a well-known class of probabilistic graphical models known as SAGA algorithms. These models, that are usually based on Hidden Markov Models or Dynamic Bayesian Networks, are essentially clustering algorithms for sequence data (genomic and epigenomic data). Similar to most unsupervised learning algorithms, evaluation of their prediction performance is challenging. In this research project, I use statistical principles of reproducibility analysis to evaluate these probabilistic machine learning algorithms.

DRUG RESISTANCE PREDICTION USING MACHINE LEARNING

2020 - 2021

github.com/mehdiforoozandeh/DRML

Deep Learning; Bayesian Optimization; Drug Resistance

In a series of related projects, I designed and optimized several machine learning and deep learning frameworks to understand and predict anti-microbial drug resistance from genomic data. In this project, I explored various avenues such as customized feature extraction from high dimensional genomic data, comparing several machine learning algorithms for the same prediction problem, and most importantly, implementing Bayesian hyper-parameter optimization to optimize our models more effectively. Particularly, I applied Bayesian hyper parameter optimization to optimize a massive deep learning model that could not possibly be optimized using conventional grid-searching method.

PREIDCTION OF ENZYMATIC PROPERTIES USING MACHINE LEARNING AND ENZYME MINING FROM METAGENOME

2019-2020

github.com/mehdiforoozandeh/MeTarEnz

github.com/mehdiforoozandeh/MCIC

https://github.com/mehdiforoozandeh/TAXyl

Enzyme mining; Metagenomics; Machine Learning; Automation

Research on developing machine learning and deep learning models for predicting enzyme properties heavily relying on amino acid sequence data for model training and feature extraction. I also designed and developed automated enzyme mining software to facilitate discovery of biocatalysts from metagenome data. Finally, using the developed computational infrastructures, we discovered and mined various novel enzymes with particular applications from metagenomic sources.

Mining Novel Industrial Enzymes from Metagnomic Sources

2019-2020

github.com/mehdiforoozandeh/MeTarEnz

github.com/mehdiforoozandeh/MCIC

https://github.com/mehdiforoozandeh/TAXyl

Enzyme mining; Metagenomics;

Research on the discovery of novel enzymes with particular applications from metagenomic sources

Papers:

Maleki, Morteza, et al. ”A novel thermostable cellulase cocktail enhances lignocellulosic bioconversion and biorefining in a broad range of pH.” International journal of biological macromolecules 154 (2020) : 349-360.

Ariaeenejad, Shohreh, et al. ”A novel high performance in-silico screened metagenome-derived alkali-thermostable endo-β-1, 4-glucanase for lignocellulosic biomass hydrolysis in the harsh conditions.” BMC biotechnology 20.1 (2020) :