MACUB (2021) Conference

Abstract: The advent of artificial intelligence (AI) found many applications of different techniques to data science and other fields including finance, engineering, medicine, physics, chemistry, biology and so on. A subfield of AI called machine learning (ML) is the first go-to place to find powerful and yet easy to implement to make multi-variate analysis where the input data consist of many properties of each sample in the dataset of your interest. In this presentation the results of application of several well-established algorithms in ML to diagnose breast cancer using characteristics/features of suspicious cells. With the diagnosis and early detection of breast cancer the 5-year survival rate of many patients increases. One method of detection of breast cancer is to extract cells from suspicious lump in the patient’s breast using Fine Needle Aspiration (FNA) technique, and to look at characteristic of individual cells or cell nuclei. This method is not as invasive as the standard biopsy that requires a surgery. A small dataset is publicly available with 30 features of sample cells (malignant and benign). Early data analyses of such data showed mixed results, depending on examiners’ skills. Among studies with the dataset, most of them ignored proper error analysis for the small statistics of the data of malignant and benign cells (212 vs. 357 samples, respectively). A rapid progress has been made in the past decade in the field of ML together with a steady increase in computational power. Thus, time is ripe to apply ML algorithms to distinguish two classes of the cells, malignant vs. benign, without human intervention to maintain consistency and good accuracy of the method. For this study we explore several ML algorithms such as Random Forest, Adaboost, XGBoost, and Support Vector Machine (SVM) to demonstrate the power of ML methods with proper error estimate on some metrics to evaluate effectiveness of the algorithms and of the FNAtechnique to diagnose breast cancer. To evaluate the metrics and their errors for performance, we adapted nested-cross-validation method that is appropriate for analysis of small data samples in this type of study. Furthermore, to avoid some bias due to population imbalance, we compare the algorithms using a balanced population from the original dataset as well as the original dataset itself. Although we found some differences among the algorithms, most of algorithms perform well with the average probability of identifying malignant cells at 94.5% and the average probability of mis-identifying benign cells at 2.2%.

Abstract: The COVID-19 pandemic has prompted the current health system to reorganize and rethink the care offered by health establishments. Breast cancer is the leading cause of cancer-related mortality worldwide, with New York having one of the highest incidences of cases in the United States. Despite recent advances in breast cancer care, the pandemic left many medical facilities changing diagnostic and treatment procedures and frequencies. This study examined changes in breast cancer care (BCC) during the pandemic. A cross-sectional analysis was done for 200+ participants in NY. Surveys were sent out to breast cancer organizations and advertised on social media. Women ages 18 and over currently undergoing treatment or in remission for breast cancer completed the survey. Answers were recorded for analysis, and responses were analyzed using a word-emotion association lexicon to assess how patients felt regarding their BCC during the pandemic. Additionally, responses were analyzed using SAS-software, and we included results corresponding with a 95% confidence interval. Our results show a total of 98.7% of patients with breast cancer received cancer treatment despite the ongoing pandemic. Most patients with breast cancer were anxious and/or worried concerning their BCC during the COVID-19 lockdown. There was no delay of appointments seen in 81.3% of patients in remission, while 15.2% of patients had a delay due to appointment unavailability and wait time after previous treatment. A total of 94.4% of patients stated that remission care has not changed since before the pandemic. However, most patients felt unsettled and nervous about the possibility of postponed treatments or canceled appointments due to the pandemic. In this study, patients with breast cancer experienced anxiety with their cancer care. The majority of patients did not experience delay and/or a reduction in the level of care required. The patient responses evaluate how the pandemic has changed BCC in NY. Our data suggest that although patients expressed anxiety and worries, care remained relatively unchanged. Regardless, BCC during a pandemic calls for policies to support and resources to identify women in need of treatment. Future work is needed to understand the full impact of the pandemic on the quality of BCC for patients across the globe.

Abstract: Emerging viral diseases have increased in recent decades. In December 2019, an epidemic with low respiratory infections emerged in Wuhan, China. The disease, Covid-19 was found to be caused by a novel coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). As of October 14, 2021, WHO has confirmed greater than 239,000,000 global cases, nearly 5 million deaths worldwide, including greater than 720,000 in the USA. Fortunately, a vaccine has been approved and distributed, yet there are no approved therapeutics for infected individuals, and the threat of emerging vaccine-resistant strains remain. From advances in biotechnology, the genome and structure of SARS-CoV-2 is known. Three proteins are anchored in the viral envelope, Spike (S), Envelope (E), and Membrane (M), which is linked to the Nucleocapsid (N) protein connecting to the viral RNA genome. Our lab is developing an innovative therapy that delivers multiple therapeutic microRNAs to block the expression of distinct Covid viral proteins. The design of the anti-Covid microRNAs 1) mimics human microRNA cluster 17-92a structural stability, 2) forms guide-RNA substrates for the RNA induced silencing complex, and 3) are complementary to specific regions of the SARS-CoV-2 RNA genome without off-targets effects in the human genome. Twenty-one microRNA sequences were designed to target the S gene, six for N, two for M, and one for E. These were cloned into our microRNA-17-92 therapy vector which expresses six distinct anti-Covid RNA therapeutics simultaneously. We have transfected the S gene into our tissue culture model to measure the efficacy of the anti-Covid microRNA therapy to down-regulate the S gene expression. In our preliminary experiments we show a significant 2.5-fold reduction in the relative abundance of the Spike mRNA in the treated cells (p < 0.05). We are currently testing additional therapies and verifying changes in spike protein levels. Next steps are to examine the secondary structure of our RNA therapy using SHAPE-MAP to optimize RNA therapeutic stability in comparison to the stable structure based on the human gene, microRNA Cluster 17-92a.

Abstract: With the diagnosis and early detection of breast cancer the 5-year survival rate of many patients increases. One method of detection of breast cancer is to extract cells from suspicious lump in the patient’s breast using Fine Needle Aspiration (FNA) technique, and to look at characteristic of individual cells or cell nuclei. This method is not as invasive as the standard biopsy that requires a surgery. A small dataset is publicly available with 30 features of sample cells (malignant and benign). Early data analyses of such data showed mixed results, depending on examiners’ skills. Among studies with the dataset, most of them ignored proper error analysis for the small statistics of the data as well as a bias due to imbalanced mixture of the data of malignant and benign cells (212 vs. 357 samples, respectively). We explore the use of modern machine learning (ML) to diagnose whether given cells are malignant or benign without human intervention. For this study we chose one of ML algorithms called Support Vector Machine (SVM) to demonstrate the power of ML methods with proper error estimate on some metrics to evaluate effectiveness of the algorithm and of the FNA technique to diagnose breast cancer. To evaluate the metrics and their errors for performance, we adapted nested-cross-validation method that is appropriate for analysis of small data samples in this type of study. Our emphasis is on the effect of population/class imbalance in the dataset. In our previous work done in 2020, we studied the effects of the population imbalance both in the training and the test dataset at the same time. In this study, we keep the population ratio constant in the training dataset, while changing the population ratio in the test dataset. This is more relevant in practice, as in reality the practitioners want to use the fixed training dataset to train their algorithm together with new test datasets collected by them without the category information (malignant vs. benign) and the population ratio. With SVM we find that over a wide range of the population imbalance it can achieve the sensitivity of 93% with 0.6 % of an uncertainty to correctly identify malignant cells, while only 2.0% with an uncertainty of 1.5% of benign cells were mis-identified as malignant. The practitioners can choose a better sensitivity value with a slight increase in the probability of mis-identification of the benign cells. We also find some biases imposed by imbalanced data and will present the result.

Abstract: Epidermal growth factor receptor (EGFR) is dysregulated in 57% of patients suffering from Glioblastoma Multiforme (GBM). It is the most common central nervous system (CNS) malignancy with a median survival of only 14 months. In our lab, we are developing assays to quantify the abundance of EGFR using quantitative PCR and DNA and RNA sequencing technologies. In order to quantify the absolute value of EGFR samples using in vitro and in vivo studies, I have isolated cDNA clones of pre-mRNA and mRNA transcripts of the EGFR gene. I isolated total RNA from either the nucleus or the whole cell, reverse transcribed with gene specific primers to produce EGFR specific cDNA, which was PCR amplified and cloned into a vector through TA cloning, the sequences were verified using Sanger sequencing to verify that EGFR is present in the cloned vectors. I am currently testing the vectors in quantitative PCR using a serial dilution to determine how effective of a standard my clones are to quantify EGFR in comparison to isolated EGFR mRNA from tissue culture of HEK 293 cells and GBM tumor cells.

Abstract:

Mesenchymal stem cells (MSC) possess paracrine activities which induce kidney repair. We aimed to assess the effects of MSC-conditioned medium (MSCcm) on diabetic kidney disease (DKD) injury in vivo and in vitro.

Male mice (10-11-weeks-old; NOD/SCID; optimal for testing xenogenic MSC therapy) were divided into 3 groups: 1) control (n=4), 2) streptozotocin (STZ 37 mg/kg i.p. for 4 days; n=3), and 3) STZ+MSCcm (n=2). MSCcm from human adipose tissue-derived MSC (7.4x106 MSC) was administered daily i.p. for 5 days. Mice were euthanized 7 days after MSCcm treatment. Ex vivo kidney tissue was assessed via qPCR for pro-fibrotic (collagen 1, activin A), pro-inflammatory (MCP-1), and kidney injury (KIM-1) markers. Fresh kidney tissue was stained for senescence marker, senescence-associated β-galactosidase (SABG). Macrophage marker F4/80 was analyzed by flow cytometry. In human renal tubule epithelial cell (HK-2) studies, HK-2 were incubated in high glucose (HG; 25mM) plus indoxyl sulfate (IS; 1mM) for 12 hours, to simulate DKD injury. HK-2 were then incubated with MSCcm for 48 hours. qPCR measured KIM-1 mRNA.

At 2 weeks following STZ injection, glucose levels (> 250 mg/dL) established diabetes in all mice. STZ induced an increase in collagen I, activin A, MCP-1, and KIM-1 which were reduced by STZ+MSCcm treatment (Figure). Decreases in cellular senescence abundance (blue SABG staining) and F4/80 (10%) were observed in STZ+MSCcm vs. STZ mice. In vitro studies confirmed a fall in HG+IS HK-2 mRNA KIM-1 expression after MSCcm treatment. In conclusion, these findings indicate a potential therapeutic role for MSC-derived cell-free therapy in DKD.  

Abstract: Anxiety is an issue that has long beset the United States for decades. From economic productivity to familial life, the nature of worrying and stress over time has changed, and as discovered in the last three decades, has become less controversial in American culture. During this national reorientation with mental health, many questions in the medical field have emerged such as: when the onset of anxiety is likely to develop, the duration of these symptoms, and what specific stimuli and conditions are correlated with higher stress levels among the population. By utilizing Integrated Public Use Microdata Series (IPUMS) Health’s database from 2000 to 2018 among 1,762,659 Americans, this two-fold observational study was able to break down anxiety indicators of Worry Frequency against the effects of one’s age. The latter half of the study observes the compounded effects of the former’s test with socioeconomic status, workloads, race, and gender. This presented several trends in how anxiety originates among the American population. In conjecturing that anxiety would increase as one ages, the regressionary analysis yielded results that indicate worry frequency on age alone is of a significant, positive relationship. Further studies provide greater clarity to the regression’s explanatory power, where in regressing aforementioned control variables, there too, was a significant and positive relationship that also exposed the existence of omitted variable bias in the first iteration. From these results, this investigation finds that anxiety does tend to increase as one ages; however, these factors are variable across people and relative to the control variables considered. Moving forward, this study is useful in applications within America to target groups that are most vulnerable to anxiety. Following this, recommendations propose that states ought to expand treatment options, counseling, and education for the general public in the near future.