To not miss a post like this, sign up for my newsletter to learn computational biology and bioinformatics.
My regret is not learning linear algebra well during college.
I barely passed the exam for it (and calculus, it was a nightmare :) ).
To be fair..
It was not taught well and it sounded too boring. I did not know what the application of matrix multiplication was, not until…
Many years later, I started to learn bioinformatics. I find many data are just data matrices:
- an RNAseq expression matrix is a gene-by-sample matrix, with entries to be read counts for each gene
- a single-cell expression matrix is a gene-by-cell matrix, with entries to be read counts for each gene
- a ChIP-seq count matrix is a peak-by-sample matrix, with entries to be the number of reads in each peak
- a drug response matrix is a drug-by-sample matrix, with entries to be IC50 for example
and many more… in other words,
Matrix is EVERYWHERE for bioinformatics (and many other data science topics)!
Many of the bioinformatics problems can be rephrased as matrix manipulation.
Read this blog post on Matrix factorization on single-cell RNAseq data to see how useful it is!
I have written two blog posts on how Seurat PCA projection/CCA alignment and label transfer work for single-cell RNAseq data in low-level details.
Man, it was hard. I spent 8 hours for each post spanning serveral nights. But I enjoyed it as I also learned the topic deeper.
During the writing of the posts, I really wished I learned linear algebra better during college:) but it is never too late to learn. You can take MIT1806, which is a great course for linear algebra.
And I am re-watching eigenvalues and eigenvectors from 3blue1brown.
You know what? Deep learning is also closely related with matrix calculations, so understand it is definitely helpful.
What’s your regret?
Happy Learning!
Tommy