For the Machine Learning Seminar, we will provide an optional activity for you to complete while the course is ongoing. This activity could theoretically be completed by those of all coding backgrounds and will be available in both Python and R. Within this activity, we will provide 3 real-life bioinformatics-based datasets (1 for classification and 2 for regression) where you will be able to learn how to go through the entire machine learning pipeline of cleaning/preprocessing your data, performing exploratory data analysis, splitting your data into training/testing sets, choosing your model(s), training your model(s), predicting on the test set, choosing the correct metric to use, and, most importantly, explaining what your results mean.

As a guide, there will be short coding snippets and guided questions to help push you in the right direction if you’re stuck along with links to online resources if you are in need of more assistance. Additionally, we plan to have short, hour-long meetings every other Wednesday (location and time TBA) where students can meet/discuss in person and where members of the Division of Applied Bioinformatics will be available for individual help. If you are unable to attend these Wednesday meetings or if you need more urgent assistance, these four Division of Applied Bioinformatics members (Roma Kurilov, Ralitsa Langova, Nicholas Abad, and Oscar Velasco Gonzalez) will also be available via email.

At the end of the Machine Learning Seminar, we are also planning a final meeting (date, location and time TBA) where those students who have participated will come to talk about their results, what they liked, what they found difficult, and most importantly discuss with their classmates why certain methods worked/did not work. No presentation will be required and we envision this to be a more open discussion rather than an actual presentation. We highly encourage all participants to come to this!

With this activity, our goals are the following:

• Implement the machine learning algorithms (i.e. linear regression, logistic regression, neural networks, etc.) that you have learned throughout the seminar using either Python and/or R
• Use real bioinformatics-based datasets to perform your analysis
• Perform an entire machine learning analysis starting from having only data available all the way to presenting your results in a clear and concise manner
• Collaborate with other students to see what works and/or does not work

• Nicholas Abad (PhD Student): [email protected]
• Ralitsa Langova (PhD Student): [email protected]
• Dr. Oscar Gonzalez Velasco (Post-Doc): [email protected]
• Dr. Roma Kurilov (Post-Doc): [email protected]

Before doing any work, it's extremely common for machine learning engineers, data scientists, and bioinformaticians to use some type of version control. Without getting into too much details, it's a way for users to "load in" someone's previous coding repository so that they can make changes or use it as a default starting point. In our case, we will be using GitHub but there are other alternatives such as GitLab, Bitbucket, Google Cloud Source Repositories, and AWS CodeCommit but thankfully, they all work the same way. If you are unaware of what version control does, feel free to watch this video: https://www.youtube.com/watch?v=USjZcfj8yxE&ab_channel=ColtSteele

If you do not yet have Git installed, follow the directions to do so here: https://git-scm.com/downloads

Now, open your terminal in your respective operating system and ensure that git is properly installed by typing:

git --help

Different git options should be displayed such as clone, init, add, merge, etc.

Now, navigate to the proper location in your computer that you want to save your work in (i.e. /Desktop/) and clone (or load in/copy) this current repository that we have by inputting the following command:

git clone https://github.com/nicholas-abad/ml-seminar-activity.git

If everything works properly, a new folder should be created in the location where you want to save your work and this folder should be called ml-seminar-activity. Ensure that this is the case before proceeding!

If you'd like to continue doing this activity using Python, ensure that you have both Python 3.xx, pip and Jupyter Notebook installed. Before starting the installation process, pip is a way to programatically download the different ML libraries that we'll be using using such as Pandas, NumPy, SciKit Learn and Keras/Tensorflow. On the other hand, Jupyter Notebooks are a common way to run Python commands in a code-block fashion. Luckily, Jupyter Notebook can be installed via pip.

To download pip, follow the directions here: https://www.geeksforgeeks.org/download-and-install-pip-latest-version/

Once this is downloaded properly, install Jupyter Notebooks by inputting the following command into your terminal/command line:

pip install notebook

After this step is complete and wihtin your terminal/command line, navigate to where this repository is saved and type in jupyter notebook. If all things work out properly, a new window will pop up. Within this window, click and open the file that ends in the common jupyter notebook file ending .ipynb and follow the directions.

On the other hand, if you'd like to continue doing this activity using R, ensure that you have R installed first and foremost and an IDE such as R-Studio. When using R, to download different packages, you could download them using the command install.packages(). Rather than having an R script, a more interactive way to display your code along with comments/images is through the use of R Markdown files. If using R-Studio, R Markdown files can be loaded in directly or if you'd like to create your own, you can go to New File > R Markdown. In order to properly view your document in a common form such as HTML or PDF, you can knit your Markdown file to html or PDF, which creates the file in the chosen format and saves this to your local machine.

In order to download the datasets, unfortunately GitHub is not the greatest way to store data and therefore we have data for you to play around with in a Google Drive. In order to get access and the link, feel free to email either Nicholas Abad ([email protected]), Dr. Oscar Gonzalez Velasco ([email protected]) or Dr. Roma Kurilov ([email protected]).