{"id":599,"date":"2022-08-19T21:51:04","date_gmt":"2022-08-19T21:51:04","guid":{"rendered":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/?page_id=599"},"modified":"2022-10-23T03:05:11","modified_gmt":"2022-10-23T03:05:11","slug":"using-independent-datasets-to-validate-a-network-map-of-glioblastoma-multiforme","status":"publish","type":"page","link":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/using-independent-datasets-to-validate-a-network-map-of-glioblastoma-multiforme\/","title":{"rendered":"Using Independent Datasets to Validate a Network Map of Glioblastoma Multiforme"},"content":{"rendered":"<div id=\"pl-599\"  class=\"panel-layout\" ><div id=\"pg-599-0\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-0\" ><div id=\"pgc-599-0-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-0-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"0\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-1\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-1-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-1-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"1\" ><div class=\"panel-widget-style panel-widget-style-for-599-1-0-0\" >\t\t\t<div class=\"textwidget\"><p>Joaquin Orozco and Teoman Toprak worked on validating a causal mechanistic flow model of the biological pathways of Glioblastoma Multiforme. This is their presentation on the topic<\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-2\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-2\" ><div id=\"pgc-599-2-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-2-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"2\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-3\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-3-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-3-0-0\" class=\"so-panel widget widget_media_video panel-first-child panel-last-child\" data-index=\"3\" ><div style=\"width:100%;\" class=\"wp-video\"><!--[if lt IE 9]><script>document.createElement('video');<\/script><![endif]-->\n<video class=\"wp-video-shortcode\" id=\"video-599-1\" preload=\"metadata\" controls=\"controls\"><source type=\"video\/mp4\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/GBM_JT_internshipvideo-2.mp4?_=1\" \/><source type=\"video\/mp4\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/GBM_JT_internshipvideo-2.mp4?_=1\" \/><a href=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/GBM_JT_internshipvideo-2.mp4\">https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/GBM_JT_internshipvideo-2.mp4<\/a><\/video><\/div><\/div><\/div><\/div><div id=\"pg-599-4\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-4\" ><div id=\"pgc-599-4-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-4-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"4\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-5\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-5-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-5-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"5\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-1.jpg\" class=\"image wp-image-614  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-1.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-1-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-1-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-6\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-6-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-6-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"6\" ><div class=\"panel-widget-style panel-widget-style-for-599-6-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">Glioblastoma Multiforme, or GBM, is a dangerous form of primary brain cancer mainly because they are aggressive and hard to remove. The World Health Organization (WHO) identifies GBM as a grade IV astrocytoma. It mainly targets the older population as the median age of diagnosis is around 64 years. Not only this, but it accounts for almost 48% of all malignant brain tumors, making it the most common malignant brain tumor. Despite it being around for a long time and being so common, there are only 5 FDA approved drugs and one device for GBM treatment since the 1920s because of its difficult-to-study nature. <\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-7\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-7\" ><div id=\"pgc-599-7-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-7-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"7\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-8\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-8-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-8-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"8\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-2.jpg\" class=\"image wp-image-613  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-2.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-2-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-2-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-9\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-9-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-9-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"9\" ><div class=\"panel-widget-style panel-widget-style-for-599-9-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">GBM is very difficult to diagnose and treat. Thus, its survival rates are extremely low at about 6.8% for 5 years. Especially for the older population which GBM mostly targets, the survival rates are a lot lower because survival rate decreases as the patient becomes older. Furthermore, the average survival from diagnosis is about 14 months which is not very long compared to other forms of cancer. Many of the most commonly diagnosed cancers have ten-year survival of 50% or more. Despite GBM being so dangerous, the methods for cancer treatment is still applicable to GBM including surgery, radiotherapy, and chemotherapy.<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-10\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-10\" ><div id=\"pgc-599-10-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-10-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"10\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-11\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-11-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-11-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"11\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-3.jpg\" class=\"image wp-image-612  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-3.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-3-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-3-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-12\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-12-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-12-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"12\" ><div class=\"panel-widget-style panel-widget-style-for-599-12-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">As glioblastoma multiforme is a troublesome form of primary brain cancer, there are challenges that come with it. First, it has heterogeneity which means each tumor can show distinctive features on a patient to patient basis and even in a given patient, there may be inter-tumor differences. There are multiple levels to heterogeneity. Next, there is a lack of clear biomarkers due to it not being studied enough. Frequent recurrence is also an issue which makes further treatment more difficult because the recurrence develops resilience. Finally, GBM situates in regions of the brain that are too dangerous to biopsy and study further. <\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-13\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-13\" ><div id=\"pgc-599-13-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-13-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"13\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-14\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-14-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-14-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"14\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-4.jpg\" class=\"image wp-image-611  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-4.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-4-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-4-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-15\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-15-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-15-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"15\" ><div class=\"panel-widget-style panel-widget-style-for-599-15-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">Models are very helpful in explaining natural phenomena. More specifically, a validated network model is very useful for things like cancer \u2013 GBM in our case. These predictive models can help us create new approaches and therapies for cancer treatment by identifying causal and mechanistic events that underlie disease progression. Furthermore, the models can predict high and low risk regulons to determine more effective treatment strategies to counter GBM. Finally, personalized cancer therapy can eventually be developed through network mapping by matching certain treatments accordingly to a person by person basis. All in all, a validated network model for GBM is essential for developing newer treatments for this aggressive brain cancer. <\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-16\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-16\" ><div id=\"pgc-599-16-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-16-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"16\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-17\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-17-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-17-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"17\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-5.jpg\" class=\"image wp-image-610  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-5.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-5-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-5-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-18\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-18-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-18-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"18\" ><div class=\"panel-widget-style panel-widget-style-for-599-18-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">This network mapping was completed using the MINER algorithm. This MINER algorithm takes in a database of genes thought to be related to GBM, and it also takes in patient data, which consists of how long they survived after diagnosis (survival data), and also their expression data. Expression is the measure of how much one gene is being activated and producing their respective protein, so the full expression dataset is done by gathering the expression of multiple genes across many different patients. MINER takes all that in, and once it\u2019s out the other side of the pipeline, it gives inferences on how each gene is activated (causal inferences) and what that gene affects (mechanistic inferences). Based off of those inferences, it filters for the genes that actually affects the life expectancy of patients. It then builds a digital model of the biological and biochemical pathways that affect different aspects of GBM.<\/span><\/p>\n<p><span style=\"font-weight: 400\">The way it simplifies the given data is by grouping the data into larger generalizations. The first step is to generate regulons, which are groups of genes that are co-expressed, meaning they share the same \u201ctranscription factor\u201d, or activation protein, thus get activated together. The data can then be split into 2 more groupings, which are transcriptional states and transcriptional programs. A transcriptional state is a group of patients that share the same gene expression, and a transcriptional programs are groups of regulons that share similar expression patterns.<\/span><\/p>\n<p><span style=\"font-weight: 400\">Now, the original model building had already taken place when we arrived at ISB. It was our job to make sure the model was representative of the real world, and we did this by using other expression datasets generated by other, independent studies.<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-19\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-19\" ><div id=\"pgc-599-19-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-19-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"19\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-20\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-20-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-20-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"20\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-6.jpg\" class=\"image wp-image-609  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-6.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-6-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-6-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-21\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-21-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-21-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"21\" ><div class=\"panel-widget-style panel-widget-style-for-599-21-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">Initially, we had the TCGA data on-hand to complete the validation testing for the gbmMINER. However, we wanted the data to be coherent across multiple datasets to increase its reliability. As a result, we used the four extra datasets of Rembrandt, Murat (GSE7696), French, and IvyAtlas. Inside of their miner inputs, we are given gene expression data for each anonymized patient and can use this data to test for high-risk and low-risk regulons and whether they are over expressive or under. By doing this, we can validate our data by ensuring that the results are coherent.\u00a0<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-22\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-22\" ><div id=\"pgc-599-22-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-22-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"22\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-23\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-23-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-23-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"23\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-7.jpg\" class=\"image wp-image-608  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-7.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-7-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-7-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-24\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-24-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-24-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"24\" ><div class=\"panel-widget-style panel-widget-style-for-599-24-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">We are able to run network mapping and test for regulon expression by patient. We want to understand whether regulons are coherent when tested with these external datasets that were not used in the model training data and almost rebuild the model. In the network mapping, we produce the coherent members (variance explained by first principal component analysis or PCA) per dataset and later perform random permutations (simply put, changing the arrangement) of a gene set within a regulon and testing for coherence with the new dataset. For example, these two graphs were taken from the Rembrandt dataset and we can see the regulon expression per each patient (<\/span><span style=\"font-weight: 400\">color key: red = overexpressed, blue = underexpressed)<\/span><span style=\"font-weight: 400\">. On the right, we see the distribution of variance that is given by the principal component analysis (PCA) for the regulon coherence.\u00a0<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-25\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-25\" ><div id=\"pgc-599-25-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-25-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"25\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-26\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-26-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-26-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"26\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-8.jpg\" class=\"image wp-image-607  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-8.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-8-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-8-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-27\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-27-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-27-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"27\" ><div class=\"panel-widget-style panel-widget-style-for-599-27-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">Now that we were confident that the regulons generated by the previous step were coherent, we then assigned a risk value to each regulon, using the Cox Proportional Hazards Analysis Regression. Fancy terminology aside, it essentially correlates the regulon\u2019s expression levels with the patient survival data. It then generates a risk value for each regulon, which determines the \u201cimportance\u201d of a regulon in determining life expectancy. For example, if a regulon has a high risk value, then that regulon\u2019s expression level changes in a patient with GBM, then there is a heightened probability of a shortened lifespan of that patient. The Cox HR Analysis also assigns a p-value to each regulon, which is almost like a confidence value &#8211; the higher the p-value, the less confident the model is that the generated risk value is accurate.<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-28\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-28\" ><div id=\"pgc-599-28-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-28-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"28\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-29\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-29-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-29-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"29\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-9.jpg\" class=\"image wp-image-606  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-9.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-9-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-9-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-30\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-30-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-30-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"30\" ><div class=\"panel-widget-style panel-widget-style-for-599-30-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">These are the last 2 steps, in which we finally compile all of the validation datasets. But first, we assign a regulator to each of the regulons. It does this by comparing the expression level of the regulator\u2019s gene to the overall expression level of the regulon, and checks to see how they match. For example, if there is a regulator with a higher expression than normal, but the regulon doesn&#8217;t change, then it is likely that the regulator does not affect the regulon, so the script just picks a new one at random and checks it again. Once, that\u2019s done running, it then compares that result to an independent analysis called the Spearman\u2019s coefficient, and if the two analyses agree, then the regulator matches with the regulon. This step is done mainly to build a model of how the regulons get affected by their regulators, as any changes in their regulator\u2019s expression can result in changes in the regulon\u2019s expression, and thus changes in the patients lifespan, for example.<\/span><\/p>\n<p><span style=\"font-weight: 400\">Then &#8211; finally &#8211; we can compile all of the datasets and look for the ones that are the highest risk in all of the datasets. We find similarly risk-valued and expressed regulons, and if all 5 datasets agree that this is a high risk regulon, then all 5 datasets must label that regulon as a statistically significant disease relevant regulon.<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-31\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-31\" ><div id=\"pgc-599-31-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-31-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"31\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-32\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-32-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-32-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"32\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-10.jpg\" class=\"image wp-image-605  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-10.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-10-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-10-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-33\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-33-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-33-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"33\" ><div class=\"panel-widget-style panel-widget-style-for-599-33-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">The causal mechanistic filtering is important because it essentially separates the disease relevant results (that are statistically significant) from the overall results and allows us to draw conclusions for just the filtered MINER results. Firstly, we use the TF-regulon statistical test for filtering results that have a PCA explained variance of &gt;=0.3 and a random permutation p value of &lt;=0.05 in both the training set (TCGA) and one of the 4 independent datasets. Only the ones with correlation and are significant are kept from the regulons. Next, we look at the Cox HR results for the kept regulons. Cox HR investigates the association between the survival time of patients and one or more predictor variables. We filter only the ones that had a Cox HR value of &lt;= 0.05 for the training data and one of the 3 datasets that had survival data (Rembrandt, Murat, or French because IvyAtlas has no survival data). We also test for significance between the mutation-eigen gene and the mutation regulator t test. Finally, all this data is compiled into a single CSV file where there is only the disease relevant C-M flows that are significant from the C-M filtering along with the survival significance from the Cox HR.\u00a0<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-34\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-34\" ><div id=\"pgc-599-34-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-34-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"34\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-35\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-35-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-35-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"35\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-11.jpg\" class=\"image wp-image-604  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-11.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-11-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-11-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-36\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-36-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-36-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"36\" ><div class=\"panel-widget-style panel-widget-style-for-599-36-0-0\" >\t\t\t<div class=\"textwidget\"><p><span style=\"font-weight: 400\">So, why is this important? Well, we are able to create maps for program risks and test for which regulons are high or low risk. For example on the bottom left graph, we can see it is a program risk stratification for every program and tests for survival probability versus the time (in days) that pass. As mentioned before, we can see trends for overexpressed regulons shown as the red trend line, along with the underexpressed regulons shown as the blue trend line. We can see that until around 500 days, the overexpressed and underexpressed regulons are at almost the same survival probability. However, we see that after 500 days there is a huge gap between the two trend lines, such that the overexpressed regulons have a much lower survival probability than the underexpressed. As a result, we can conclude the overexpressed regulons are high risk due to a lower survival probability. Furthermore, we will use patient survival data to create heatmap models regarding the individual\u2019s programs and indicators showing whether the genes are over or under expressive. Since every patient is different, we gain insight into the behaviors of each specific program by a patient basis.\u00a0<\/span><\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-37\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-37\" ><div id=\"pgc-599-37-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-37-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"37\" ><\/div><\/div><\/div><\/div><div id=\"pg-599-38\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-38-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-38-0-0\" class=\"so-panel widget widget_media_image panel-first-child panel-last-child\" data-index=\"38\" ><img loading=\"lazy\" decoding=\"async\" width=\"960\" height=\"540\" src=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-12.jpg\" class=\"image wp-image-603  attachment-full size-full\" alt=\"\" style=\"max-width: 100%; height: auto;\" srcset=\"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-12.jpg 960w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-12-300x169.jpg 300w, https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-content\/uploads\/sites\/10\/2022\/08\/JT_GBM_Internship_Presentation-12-768x432.jpg 768w\" sizes=\"auto, (max-width: 960px) 100vw, 960px\" \/><\/div><\/div><\/div><div id=\"pg-599-39\"  class=\"panel-grid panel-no-style\" ><div id=\"pgc-599-39-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-39-0-0\" class=\"so-panel widget widget_text panel-first-child panel-last-child\" data-index=\"39\" ><div class=\"panel-widget-style panel-widget-style-for-599-39-0-0\" >\t\t\t<div class=\"textwidget\"><p>So what are the next steps for this project? One option is to infer and and understand any existing and novel biological pathways that the model infers. This is a good example of using the systems biology approach: considering an element&#8217;s interactions with other elements within its system to generate a more accurate model for real world applications. Another next step could be combining all of the Jupyter notebooks we used in the project into one single notebook or python file. This would give ISB&#8217;s work a greater outreach and make it accessible to more people who have less of a data science or computational biology background. And there is another direction the project could go in: expanding the use of the MINER algorithm, and other algorithms like it, to more complex diseases. Since MINER has already been used in multiple myeloma and now GBM, it would not be hard to expand its usage to other, more complex diseases that we have yet to understand.<\/p>\n<\/div>\n\t\t<\/div><\/div><\/div><\/div><div id=\"pg-599-40\"  class=\"panel-grid panel-has-style\" ><div class=\"panel-row-style panel-row-style-for-599-40\" ><div id=\"pgc-599-40-0\"  class=\"panel-grid-cell\" ><div id=\"panel-599-40-0-0\" class=\"so-panel widget widget_block panel-first-child panel-last-child\" data-index=\"40\" ><\/div><\/div><\/div><\/div><\/div>","protected":false},"excerpt":{"rendered":"<p>Joaquin Orozco and Teoman Toprak worked on validating a causal mechanistic flow model of the biological pathways of Glioblastoma Multiforme. This is their presentation on the topic<\/p>\n","protected":false},"author":76,"featured_media":448,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-599","page","type-page","status-publish","has-post-thumbnail"],"_links":{"self":[{"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/pages\/599","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/users\/76"}],"replies":[{"embeddable":true,"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/comments?post=599"}],"version-history":[{"count":9,"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/pages\/599\/revisions"}],"predecessor-version":[{"id":1110,"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/pages\/599\/revisions\/1110"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/media\/448"}],"wp:attachment":[{"href":"https:\/\/baliga.systemsbiology.net\/see-interns\/hs2022\/wp-json\/wp\/v2\/media?parent=599"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}