Monday: history
I’m a 3rd year PhD student (2021) at the University of Birmingham, UK. My field is Clinical Informatics, and I am looking at inflammation! Today I’ll talk about my experience learning to code throughout education: when it started, how it progressed, and what I can do now!
I was lucky to attend a primary school with a computer suite. My first search engine was Ask Jeeves. I didn’t understand what a search engine was at the time, seeing “Ask” meant I would literally ask it full questions! These days Google already knows what I want to ask before I have finished typing.
I was around 8 years old when I first encountered coding. A school visitor introduced us to “drag & drop” visual programming with a tool called Flowol (like Scratch). We started with a few exercises: arranging shapes to make a sequence of commands. I remember enjoying it so much that I rushed through to ensure I could attempt all the exercises.
We finished with a competition: groups would race to solve the traffic light problem and students had to consider the rules. My partner and I struggled but I persisted. I remember my sequence of shapes became very complex at the end with arrows pointing everywhere, but it worked!
Our school also had a programmable floor robot. There was only one for the school so we couldn’t individually try. The teacher would show us how to run it. As much as I enjoyed these experiences, I wouldn’t cross paths with programming again until I was in college.
In college (aged 16) I decided to study IT as realised I was actually pretty good at it (plus I enjoyed it too). This course covered many areas of computing, including: animation, game design, business, and web development. I still have the animation project available to anyone to watch - Little Lonely Bird - the aim was the provoke emotion.
I really dived in with web lessons. We had to build a basic website with HTML and CSS with Adobe Dreamweaver.
I didn’t have a computer at home, so I spent my evenings at school to ensure I wouldn’t fall behind.
There was another web course which included some Javascript - the teacher provided some code snippets, which I further explored in my own time.
I vageuly recall adding a clock to my site.
When I got a laptop of my own, I tried to learn Javascript in my spare time.
It was at this stage that I had to start thinking about my future: what were my career options? I found IT “easy” and fun. I asked my IT teacher for advice, who strongly encouraged me to do Computer Science and told me to choose a university outside my hometown to gain confidence.
I started my undergraduate degree at Aberystwyth University, UK in 2014.
Some coding skills I learnt in my first year: Arduino C and Java.
I struggled with Java as going from Arduino C to an object-orientated language was too confusing.
But with every module, I took a book from the library to try and understand more.
One of my first challenges as an undergrad was the introduction to Linux and the command line interface: this was stepping into a whole new computing world.
It was nothing like I had ever done before.
My second year modules included: web programming, system administration, and databases.
New skills gained included: PHP and PostgreSQL.
My third year included a group project and my dissertation.
The disseration was a major web project: I developed a site for a small cake business with a database.
This is also the year I was introduced to Machine learning which we used the Weka software.
At the end of my third year, I started to think about my next steps: I decided to apply for a Masters in for Data Science.
During the Summer of 2017 - after undergraduate graduation and before my Masters - I started a research position in bioinformatics.
This is also the Summer I really started to dive in Python - with help of course!
This is where I finally figured out programming: at all clicked.
Autumn 2017 was the start of my Masters in Data Science!
I explored various topics and dived deeper into Machine learning, databases, and newly statistics.
This expanded my skill set with: R and MongoDB.
One of my most memorable and favourite university modules was the Python module.
This was taught through “live” coding sessions, which the lecturer would get us involved.
My Masters disseratation project used a variety of skills,
The dissertation involved looking at Acidobacteria and its DNA content.
Here I really explored my Python skill and developed my first Python tool and released it on GitHub.
Tuesday: PhD
PhD applications: over the years, I’ve had questions from students about PhD applications, including essays & interviews. Note to reader: although I Tweeted advice, to avoid duplication you read the information in my PhD applications blog.
My original PhD title was different: the plan was to use various Biobanks and link metadata gaps, however, we soon learnt that the data wasn’t available…
Creating an AI based data assistant to bridge genotype to metadata linking primary clinical data to biobank sample
The new aim is to investigate inflammation using the UK Biobank. I am looking into non-traditional biomarkers of inflammation through various methods & data sources: structured and unstructured. Some non-traditional biomarkers include blood assays and clinical letters.
Note to reader: this takeover was in 2021 (3rd year) and I have since completed my PhD! Readers can read about it here.
To link back to yesterday about coding experience, currently my main language is Python but I also use R for statistics or making nice plots.
Wednesday: skills
My main interests are: Machine learning, ontologies, and natural language processing!
Machine learning (ML) is “learning from data”, it can be supervised prediction with labelled data or unsupervised clustering to reveal patterns in data.
I’m most interested in clustering and have experience using K-means, hierarchical, DBScan, spectral, and more!
Don’t forget visualisation - it’s important to see what the data looks like.
To visualise data with a high number of dimensions, I explored various dimensionality reduction methods: specifically t-SNE and PCA. An interesting tool I found is PCAmixdata (in R) essentially combining PCA (continuous) and MCA (categorical).
For all things Machine learning, I use the scikit-learn module in Python which also does statistics, dimensionality reduction, and methods for the optimal number of K.
I like to say that there’s no “right way” to do Machine learning - but parameter finetuning can improve results!
Ontologies condense a domain of knowledge in formalised structure. The concepts of an ontology have metadata and relationships. For example, in human anatomy: hand “part of” arm [synonym = upper limb].
Ontologies help with semantic similarity: allowing us to do association text mining: extracting important information from a document using terms and corresponding synonyms. In relation to my PhD: I aim to look at word vectorisation of clinical letters: looking how inflammatory terms are closely related to one-another.
My work in natural language processing (NLP) has proved difficult as many tools don’t allow for easy ontology use and so additional wrangling is needed… A first step to overcome this barrier: I developed Jabberwocky, an NLP toolkit for the easy manipulation of ontologies.
If you are interested in NLP, I recommend spaCy as it has a brilliant tutorial/guide.
Moreover, if you are interested in visualising an ontology, I recommend the WebVOWL online tool.
Note to reader: I’ve found WebVOWL to not be 100% reliable. I have since developed Jabberwocky eyes() function for plotting ontologies.
Finally, I have also been dipping my toes into the genomics side of bioinformatics via genome-wide association studies (GWAS): looking into genetic variants of a group of people to observe common traits. To be honest, this is not my strongest skill: I am not confident in genetics, but others in my lab created a script and shared with me to help!
Thursday: struggles
One of my biggest struggles is trying to communicate my work - I’m still learning about the correct terms I should be using and how to present my results… An example of this is: using “feature” instead of “column” when describing data.
Futhermore, I’m quite nervous about conferences because I worry that I’ll be asked questions that I can’t answer (relating to my struggle in communicating). In all though, conferences are good for networking and career experiences.
Continuing with the struggle in communicating my work…Although I enjoy writing, my grammar is not the best - BUT it doesn’t stop me from trying. It started off quite difficult to accept feedback as it came from a place of embarrassment. Now I value comments and I appreciate others spending time reviewing my work!
Relating to work itself, it’s important for me to have some sort of IDE while programming.
I “need to see” my variables to continue, perhaps this comes from a place of doubt as I’m still not 100% confident in my programming skills.
I use Spyder for Python and RStudio for R.
One major struggle is not specifically related to the PhD, but rather trying to survive this pandemic. I’m lucky to not have my work affected as it’s computational - but I still find it difficult. I’m too nervous to go outside because I see a lot of people not wearing masks and times are scary for me… My body aches as I don’t get my daily walk & my back hurts from my cheap chair!
It’s very important to take time for yourself and do things you enjoy! I’ve had to take a day or few off in the past year because of the overwhelming stress… No matter in academia or industry, you need to take care of yourself and businesses/companies need to take care of their employees!
The PhD is an emotional rollercoaster, in 2018 I wrote up about witnessing my partner writing their PhD and now being in that position, I completely understand!
Friday: funday
Hobbies! Since the pandemic I’ve started a bunch of hobbies: knitting, clay sculpting, and I also coding more for fun! I have a Colours project that’s been in development for a few years: small tools developed with the techniques I’ve gained throughout my years in education!
Music I listen to whilst working: I love soundtracks! Such as Tron Legacy & Interstellar… Hans Zimmer is my go-to!