Author(s): Omer Ozgur Introduction To Genomic Data Science From https://www.bezelyedergi.net/post/biyoinformatik-sekanslama-teknolojileri Every living organism has its genome at its heart. The genome sequence of a cell’s DNA is its software. We can analyze DNA with the same logic if we view it as software that runs inside the cell. The DNA storehouse is more than just information. DNA is an incredibly complex physical structure. The genomes are complex, multi-part machines that can be broken down into thousands. While we are able to understand how certain genes function today, many of the genes interact with each other. Geneticism sees DNA as only information. It looks for patterns in data and investigates the relationship between genes. Genomics, on the other hand sees the genome and attempts to figure out how it works together. Genomic Data Science Genomic Data Science uses methods from data science such as machine learning and statistics to solve genomic problems. Next-generation sequencing allows us to sequence genomes more efficiently, faster and cheaper. The Human Genome Project cost $2.7 billion. Today, you can sequence your DNA for $1000.. You can find petabytes worth of genomic data on the internet. The complexity of a living organism surpasses that of any computer. This complexity requires more than human and classical algorithmic understanding. Many genetic diseases, such as cancers and genetic disorders, are due to the interactions of multiple genes and have genetic variation. Deep Learning is one of the best tools for analysing complex data points. Traditional methods presume linear relationships between genes. Deep Learning is a powerful tool. Genomic Data Science allows personalized drug research to be done faster. If you are bored you can also create a 3D model of the person’s faces using their genomes. This can help you detect those who have used tobacco or chew gum. The principles of life and DNA are similar. The tree of life links all living beings together and is most likely derived from one common ancestor. The DNA polymer is long and contains four basic bases (A,T, G, and C). Nearly all of the information necessary to create an organism can be found here. This information and its processing (epigenetics), changes as time passes. There are almost infinite combinations of information that DNA can store. This space holds the code to unlock new life or people that were not born. The Central Dogma Proteins are as important hardware as DNA. Proteins, which are small machines inside cells that perform most of the work required to make them function properly, are the heart and soul of our cell. When converting DNA information into proteins, a molecule called the mRNA must be present. The Ribosome receives the mRNA and, depending on the information, the amino acids and proteins are then combined. Organic 3D printers called Ribosomes. Let’s now look at how genomes function. Eukaryotic DNA has been wrapped in proteins known as histones for a snug fit within the cell. Tightly packed areas are not readable and methylated portions are hard to read. We are still not sure of the mechanisms which control when DNA should open. It is not possible to have a unidirectional flow from DNA to proteins. By binding to DNA, proteins can act as regulators. The mRNA contains information about the protein being synthesized but does not specify when. Here is where transcription factors are important. These transcription factors are able to bind to DNA at specific locations and control the expression of genes nearby. Editing tasks can be performed by siRNA, miRNA and Riboswitches. It is important to note that not all editing mechanisms within the genome can be understood and the process may prove more complicated than what we have described. Predicting Transcription Factor Binding. We discovered that cells are complex and difficult to manipulate with traditional methods. Therefore, we will use deep learning techniques. As data, we will be using the JunD Transcription Factor and HepG2 cell lines. HepG2 is an immortal cell line derived from the liver tissue of a 15-year-old African-American child. JunD transcription factor encoded in the JUND gene. This factor can activate or downregulate other genes. We chose the 22nd Chromosome so that the genomic data could be processed. The Chromosome is home to about 50 millions base pairs. Humans have approximately 3 billion base pairs. The largest Chromosome is the 1st, and the smallest is the 22nd. FASTA and FASTQ formats are used to store genomic data. The data will appear like book rows. >chr22 NNNNNNNNNNNNNNNNNNNNNNNNNNN TCCCAAATTGTGGAAGGAATGTACATTTGAC Here we see the base sequences contained in a single strand of DNA. It is possible to ask the question “What does N” stand for in DNA? It cannot be determined which base will be read during the sequence process. The next step is to turn the genomic data into a format we can use. Our data is made up of four parts: A, T. G. C. and N. One-hot encode allows us to express this information. You can represent A, T, G and C with One-hot encoding. [[1., 0., 0., 0.],[1., 0., 0., 0.],[0., 1., 0., 0.][0., 0., 1., 0.],[0., 0., 0., 1.]] == [“AATGC”] After DNA one-hot is encoded, we can start training. The 101 base sequences and the bound transcription factors make up our training data. If there are 2000 base, then we have 20 row. The problem can then be supervised. Our x[0] == [[1., 0., 0., 0.],[1., 0., 0., 0.],[0., 1., 0., 0.][0., 0., 1., 0.],[0., 0., 0., 1.]].. our y[0] == 1 We can use many models to find the relationship between the x’s and y’s. CNN will be used for this example. From https://www.researchgate.net/figure/Causal-convolution-operation-in-a-1D-convolutional-layer-with-k-3-kernel-size-Input_fig1_337703712 Since DNA is a string, it is one-dimensional. There are many architectures that we can choose from. Convolutional neural networks are one of the best architectures to find patterns in 1D data. These networks are used for images (also known as 2D information), and can also be used to process text. 2D CNNs learn about the edges, vertices, colors, patterns, and other details of an image. We can also learn filters to extract information from text data. After our model has been trained, it is possible to calculate how likely the JunD protein will attach to DNA fragments 101 long. More realistic results From https://www.scinexx.de/dossierartikel/auch-die-verpackung-machts/ Chromatin accessibility: Defines how accessible the DNA is to molecules from outside. Transcription Factors and other molecules cannot be accessed if DNA is wrapped tightly around histones. Genes that need to be silenced may also be packaged. Accessibility to a particular region may change over time. This can change over time. We can predict more accurately if we include the information about chromatin accessibility to our data. Conclusion and Example code The complexity of biology goes beyond our human comprehension. Machine learning is essential for improving our understanding and performance. Scientists thought that they knew all of the secrets when the genome was sequenced. But, they were only scratching the surface. CRISPR and genome editing tools such as CRISPR are not the solution. Editing is impossible without knowing the effects of genetic mutations. If not in an unexplicable manner, unravelling the secrets of the genome would offer an opportunity for us to.
