The iResearch Institute is the leader of international science research training!

iResearch Institute 2021 Student Highlights

Elizabeth Luo

Genetics and Chemotherapy-Induced Peripheral Neuropathy
Hometown: Longmeadow, Massachusetts
Mentor: Anthony Park, Purdue University

Chemotherapy-Induced Peripheral Neuropathy (CIPN)  is a side effect of chemotherapy that can lead to symptoms ranging from numbness and pain to organ failure or paralysis. It may necessitate a reduction or cessation of chemotherapy, negatively impacting treatment. By running differential gene expression analysis (DESeq2) using RStudio, it was possible to find which genes are highly over or under-regulated in patients that develop CIPN. After comparing the mechanisms of CIPN with the significant genes (p-value 0.05), 11 genes that had a direct connection to CIPN were identified. As this research is continued into other chemotherapy side effects, such as cardiomyopathy, the information will allow better-personalized care of cancer in deciding whether or not to use chemotherapy or which chemotherapy to use.

Arjun Singh

Unsupervised Machine Learning Approach for Identifying Biomechanical Influences on Protein-Ligand Binding Affinity
Hometown: Warren, New Jersey
Mentor: James Wang, Columbia University

Drug discovery is incredibly time-consuming and expensive, and researchers have yet to uncover most of the biochemical properties that make drugs viable. In this study, Machine Learning (ML) was used to reveal which molecular structures were most responsible for viable drugs. An ML pipeline was used to group drugs into those that bind strongly and weakly to target proteins. Correlation analysis was performed to validate which drug fragments were most responsible for the grouping. It was found that the CCCH and CCCCCH molecular fragments are most responsible, revealing significant biochemical insight into the molecular binding properties of viable drugs.

Shriya Bhat

Combating Bacterial Biofilm III: Investigating Quorum Sensing Pathways in Multispecies Biofilm and Developing a Non-toxic Quorum Quenching Cocktail Therapy to Reduce Biofilm Virulence 
Hometown: Richardson, Texas
Mentor:
 Emily Gan, MIT

Dense multi-species bacterial biofilms are resistant to most antibiotics. In this study, a combination treatment consisting of FDA-approved concentrations of quorum quenching agents was devised to target three major interspecies biofilm-pathways. Computational analysis (Qiime2/PICRUSt2) of the CF lung microbiome confirmed the necessity of QS in targeted pathways, and docking analysis supported the binding affinities of each treatment and targeted enzyme (LasR/LasI/PqsR). The combination treatment demonstrated a near 80% inhibition efficacy on P.  aeruginosa-S. aureus-B. cepacia-C. albicans complex in-vitro. These findings can be translated into the development of novel adjuvants to deliver the cocktail treatment in-vivo, thus reducing mortality from chronic biofilm-related infections.

Meredith Joo

Elucidating biomarkers shared between systemic lupus erythematosus and inflammatory breast cancer
Hometown: Ridgefield, Connecticut
Mentor: Vivian Utti, Corneil University

Despite the clinical similarity between systemic lupus erythematosus (SLE) types and inflammatory breast cancer (IBC), IBC is not considered classically “inflammatory” because its inflamed appearance is caused by lymphatic blockage, and genetic characteristics common to both diseases are underexplored. This study sought to reveal those characteristics and reassess IBC’s “inflammatory” status. Shared SLE and IBC genes were found using Python and tested for pathway and transcription factor (TF) enrichment using Enrichr and ChEA3. These analyses yielded inflammatory pathways and cell cycle-related TFs in the upregulated genes, and translational pathways and cell proliferation TFs in the downregulated genes. The similarity between the diseases, in inflammation and altered translational function, potentiates the need for new treatment development.

Sriram Ayalavarapu

Ulcer Localization and Classification using Faster Region-based CNNs from Capsule Endoscopy Results
Hometown: Frisco, Texas
Mentor: James Wang, Columbia University

Gastrointestinal Ulcers are linked to diseases that result in a risk of Colorectal Cancer. As such, analyzing Capsule Endoscopies, which record the digestive tract, for GastroIntestinal Ulcers could assist in the diagnosis of Cancer. Therefore, I implemented a state-of-the-art Machine Learning Model to localize and detect ulcers with higher precision than existing models. The results show significant progress but seem to suggest issues that will be fixed through Data Augmentation. After these improvements, this model will be capable of accurately detecting Ulcers in endoscopy footage, allowing it to assist in diagnosing Crohn’s disease, Ulcerative Colitis, and Colorectal Cancer.

Nicholas Djedjos

Identifying Genetic Biomarkers for Diagnosing Essential Tremor
Hometown: Brandon, Mississippi
Mentor: Prathamesh Chati, Washington University @ St. Louis

Nearly seven million individuals in the U.S. have Essential Tremor (ET). Differential Gene Expression (DGE) on ET RNA-seq data identified 86 differentially expressed gene transcripts. The gene transcripts were input into Gene Set Enrichment Analysis (GSEA) where five dysregulated gene pathways were identified. After filtering the genes to 32 with optimization, a Random Forest model predicted ET from control 85% of the time. Logistic regression was utilized to analyze the 32 genes individually, and 8 genes had a higher accuracy than 80%. Identifying these biomarkers shows that ET may be related to other neurological diseases more than once thought.

Shriya Bhat

Combating Bacterial Biofilm III: Investigating Quorum Sensing Pathways in Multispecies Biofilm and Developing a Non-toxic Quorum Quenching Cocktail Therapy to Reduce Biofilm Virulence 
Hometown: Richardson, Texas
Mentor: Emily Gan, MIT

Dense multi-species bacterial biofilms are resistant to most antibiotics. In this study, a combination treatment consisting of FDA-approved concentrations of quorum quenching agents was devised to target three major interspecies biofilm-pathways. Computational analysis (Qiime2/PICRUSt2) of the CF lung microbiome confirmed the necessity of QS in targeted pathways, and docking analysis supported the binding affinities of each treatment and targeted enzyme (LasR/LasI/PqsR). The combination treatment demonstrated a near 80% inhibition efficacy on P.  aeruginosa-S. aureus-B. cepacia-C. albicans complex in-vitro. These findings can be translated into the development of novel adjuvants to deliver the cocktail treatment in-vivo, thus reducing mortality from chronic biofilm-related infections.

Indeever Madireddy

Using Machine Learning to Develop a Global Coral Bleaching Prediction System 
Hometown: San Jose, California
Mentor: Dr. Rachel Bosch, University of Cincinnati

Coral bleaching is a fatal process in which corals expel their symbiotic dinoflagellates. Restoration efforts have attempted to repair damaged reefs, however, there are over 360,000 square miles of coral reefs worldwide, making it challenging to target conservation efforts. Herein, a machine learning model was developed to predict global locations at risk for coral bleaching. Data was obtained from BCO-DMO and consisted of various coral bleaching events and the parameters under which the bleaching occurred. Various models and classifiers were tested and a 95% accuracy in predicting the likelihood of a bleaching event was obtained.

Jasmine Lu

Convolutional Neural Network Models for Classifying Shoulder Images of Musculoskeletal Disorders
Hometown: Taipei, Taiwan
Mentor: Sophia Ladyzhets, Columbia University

Shoulder pain is one of the most common physical injuries, as two out of three people experience this condition at some point in their lives. Typical causes of shoulder pain include biceps tendinitis, rotator cuff tear and tendinitis, and bursitis. This study aimed to use convolutional neural network (CNN) models for evaluating ultrasound images to help primary care physicians diagnose shoulder pain. The CNN models implemented in this study are DenseNet 201, Xception, and InceptionResNet. The three models had satisfactory performances that demonstrate the potential of assisting primary care physicians by increasing diagnostic accuracy and providing rapid preliminary results.

Tarunika Sasikumar

Assessing the Efficacy of the U.S Endangered Species Act through Analysis of Species Charisma and Respective Population Trends
Hometown: Plainview, New York
Mentor: Dr. Rachel Bosch, University of Cincinnati

The U.S Endangered Species Act (ESA) is legislation implemented to repopulate endangered species affected by the ongoing sixth mass extinction. In this study, the links between species charisma and population trends are investigated to assess the efficacy of the ESA from a non-economics perspective. 367 vertebrate species were ranked based on charisma (through a novel point value system) and their rankings were compared to respective qualitative population trends. Species with higher charisma levels had a greater chance of possessing a “beneficial trend” in their population levels. These findings suggest a detrimental bias, based on outward appearances, present in the ESA’s effort to repopulate at-risk species.

See 2020-2021 Highlights