BME Graduate Student Seminar (Zoom): Raji Nagalla and Christopher Toh
Raji Nagalla: Immunomodulation in wound healing
Abstract: There is a large and growing clinical need for improved wound therapies. Skin wound healing involves the orchestrated communication and activities of macrophages with other wound effectors. Wound macrophages and fibroblasts respond dynamically to changes in their local physical and biochemical environment, presenting a target for engineered biomaterials to modulate cell-cell interaction in the wound bed. This dissertation examines the reciprocal signaling between macrophages and fibroblasts, and the potential of biophysical properties of engineered hydrogels to modulate this interaction to improve wound healing. Soft gelatin methacrylate (gelMA) hydrogel was shown to reduce scar size in small, full-thickness murine skin wounds, compared to stiff gelMA and no treatment, additionally promoting a pro-healing macrophage phenotype in vitro and in vivo. Single-cell RNA sequencing of wound tissue treated with soft or stiff gelMA or no material at post wound day five revealed heterogeneous macrophage and fibroblast populations, with distinct shifts and differential gene expression in response to material stiffness. Cell-based wound closure assays in 2D and 3D were used to further parse these interactions, showing that juxtacrine coculture of murine bone marrow-derived macrophages with NIH 3T3 fibroblasts significantly enhanced fibroblast closure of 2D and 3D wounds. Coculture also altered macrophage activation in a contact-dependent manner, when compared to culture of ether cell type alone. Finally, broad inhibition of gap junctions with palmitoleic acid abrogated fibroblast-enhanced macrophage IL-10 secretion and coculture enhanced calcium activity, suggesting that cell-cell contact through gap junctions may, in part, mediate macrophage-fibroblast communication. This work demonstrates a critical role for direct macrophage-fibroblast interactions in the cellular coordination of wound healing and reveals the potential for targeting biophysical immunomodulation in the development of wound healing therapeutics.
Christopher Toh: Chromosomal scale length variation as a genetic risk score for predicting complex human diseases in large scale genomic datasets
Abstract: The advent of the digital information age and next generation sequencing has created large databases of human genomic information. Utilizing this information in order to understand disease and genetic risks is a large engineering task. Previous studies have focused primarily on single nucleotide polymorphisms (SNPs) as the primary feature of interest in assessing patient risk for diseases such as cancers and other highly heritable diseases such as schizophrenia. However, these SNP panels do not take into account epistatic interactions between various portions of the human genome.
Chromosomal scale-length variation (CSLV) is a promising new approach for assessing genetic risk scores. This method includes epistatic effects that might be missed by conventional genome wide association studies (GWAS). GWAS typically use a linear combination of SNP scores to assess genetic risk. CSLV evaluates copy number variations (CNVs) across large sections of the human genome to obtain a comprehensive account of variations that may contribute to inheritance of disease risk.
Utilizing modern machine learning classification algorithms, we assessed prediction of diseases such as ovarian cancer and schizophrenia using CSLV as the sole feature for prediction. CSLV is a measure of a person’s genetic variation using CNV as the basis for the measurement and examination of this variation across the large sections of the chromosomes. Utilizing large databases such as The Cancer Genome Atlas (TCGA) and the UK Biobank, we have demonstrated the viability of this method in assessing germ line inheritance of complex human diseases. We utilized h2o, a machine learning framework and assessed the performance of several models, including general linear models (glm), gradient boosted machines (gbm), XGBoost and stacked ensembles.
Using normal blood samples from the 11,000 patients in TCGA, we tested 33 different types of cancers. We set up a binary classification task between patients with the diagnosed cancer and those who did not have the diagnosis. Of note several cancers had high performance or the models had a large area under the curve (AUC), the primary method of assessing machine learning models. Glioblastoma multiforme (AUC = 0.87), ovarian cancer (AUC = 0.89), colon adenocarcinoma (AUC = 0.82), and breast invasive carcinoma (AUC = 0.75) could be predicted greater than chance when compared to other cancers.
We then replicated this experiment in the UK Biobank using the 88 number computed from the 22 autosomes for 1,534 women with breast cancer and a control population of 4,391 women without breast cancer and found a classifier with an AUC of 0.83. This demonstrates breast cancer can be predicted greater than chance from the general population.
Additionally, we examined other complex human diseases with evidence of genetic heritability, namely schizophrenia. We performed analysis of 1,129 people from the UK Biobank with a diagnosis of schizophrenia. Using a randomized set of 1,129 individuals without schizophrenia, we created 150 models using the 88 number CSLVs as our feature set. The results provided an average AUC of 0.583 (95% CI 0.581-0.586). Our results indicate that germ line chromosomal scale length variation data can provide an effective genetic risk score for schizophrenia. Additionally, we found that a general linearized model is often the best model, but we found that gradient boosted machines and XGBoost work just as well. A stacked ensemble when it does work provides the absolute best performance.
In conclusion, CSLV is a promising and novel way to utilize large scale human genetic information in the prediction and onset of complex diseases that show evidence of germ line inheritance. Continued improvement of this technique can dramatically improve individualized patient care and aid physicians in earlier diagnosis and preventative treatments.