Bioinformatics sits between biology, computing and statistics, and people reach it from all three directions. That flexibility is an advantage, but it also makes the path confusing: there are many tools, languages and opinions, and it is not obvious what to learn first. This roadmap is written for students with a biology, biotechnology, microbiology or pharmacy background, and is useful for computer-science students moving the other way.
What bioinformaticians actually do
Job titles vary (bioinformatician, computational biologist, bioinformatics analyst, research software engineer, data scientist in life sciences), but the day-to-day work usually combines several of these:
- Processing and analysing sequencing data: genomes, exomes, transcriptomes, single-cell and microbiome data.
- Building and maintaining reproducible pipelines that other scientists run.
- Statistical analysis and visualisation for papers and grant reports.
- Structural bioinformatics and computational chemistry in drug-discovery teams.
- Developing new methods or software, which is more common in research groups and tool-building companies.
Most roles involve more data handling and communication than students expect, and less novel algorithm design.
Stage 1: foundations (first three months)
Biology, from a data point of view. Revisit the central dogma, gene structure and genetic variation, but ask what each experiment actually measures. What does a read count represent? What does a variant call mean?
The command line. Almost every bioinformatics tool runs in Linux. Learn to move around the file system, use grep, cut, sort, awk and pipes, and install software with conda.
One programming language. Pick Python or R and stay with it until you are comfortable. Python is more general-purpose and widely used in pipelines and machine learning. R is strongest for statistics and hosts Bioconductor. You will likely use both eventually, but learning two at once slows you down.
Core formats and databases. FASTA, FASTQ, SAM/BAM, VCF and GFF; NCBI, Ensembl, UniProt and PDB.
Stage 2: statistics and one application area (months 3 to 6)
Statistics. Distributions, hypothesis testing, regression, and especially multiple-testing correction and the false discovery rate. Genomic datasets test thousands of hypotheses at once, and misunderstanding this is the most common source of wrong conclusions.
One application in depth. Choose the area closest to your interests and complete it end to end:
- Bulk RNA-seq: QC, quantification, DESeq2, enrichment.
- Variant calling: alignment, GATK or bcftools, annotation.
- Structure-based drug discovery: protein preparation, docking, validation.
- Microbiome: QIIME 2, diversity and differential abundance.
Depth matters more than breadth here. One analysis you can explain and defend is worth more than ten tutorials you followed without understanding.
Stage 3: reproducibility and portfolio (months 6 to 12)
Version control. Learn Git and put your work on GitHub. Employers look at it.
Workflows and environments. Learn the basics of a workflow manager (Nextflow or Snakemake) and of containers (Docker or Apptainer). Try running an nf-core pipeline.
Build a portfolio. Two or three well-documented projects carry more weight than a list of certificates. Good projects:
- Re-analyse a published dataset from GEO or SRA and compare your results with the paper's.
- Write a small, tested tool that solves a real problem and document it properly.
- Contribute a fix or documentation improvement to an open-source bioinformatics project.
For each project, write a short README explaining the question, the data, the methods, the results and the limitations.
Getting experience
Internships and project work. A mentored project with code review teaches habits that are hard to learn alone, and gives you someone who can describe your work in a reference.
Research groups. Many academic labs have more data than analysts. Offering to help with a well-defined analysis, under supervision, is a common route into a first role or a postgraduate position.
Communities. Biostars and the Bioconductor support site answer most technical questions; read the existing answers before asking. Following nf-core and other open-source projects is a good way to learn how professional pipelines are built.
Common mistakes
- Collecting tools instead of understanding. Knowing that DESeq2 exists is not the same as knowing when its assumptions fail.
- Skipping statistics. It is the part interviewers probe most.
- No public work. If nobody can see what you have done, it is hard for anyone to judge it.
- Waiting to feel ready. Start a real project early; it shows you what you still need to learn.
A structured path
Our learning paths follow this roadmap. The Genomics Data Scientist path runs from Bioinformatics Foundations through Python, NGS analysis, RNA-seq and single-cell analysis. The Computational Drug Discovery path covers docking, QSAR and machine learning. Both end in projects you can put in a portfolio. Our internships add mentored work on real problems.
Written by the Vivofold team for research and education. Found an error? Tell us and we will correct it.