The Pizza Yeast BLAST Surprise
BLASTing mysterious 'novel human brain gene' sequenced after 3 months of wet-lab prep.
The E-value (Expect value) describes the number of random alignments expected by chance; E-value 0.0 represents statistical certainty, often uncovering cross-contamination.
FASTA Line 2 Trailing Whitespace
Writing a 500-line Python genomics script to process 10,000 genomes.
FASTA format requires strict adherence to header lines starting with '>' followed immediately by sequence records on subsequent lines with no unexpected delimiters.
Affine Gap Penalty Budgeting
How is sequence alignment like signing an apartment lease?
Affine gap penalty models reflect the biological reality that a single multi-nucleotide insertion/deletion event is much more probable than multiple independent single-base indels.
Dynamic Programming Big-O Fear
Needleman-Wunsch & Smith-Waterman: 'We guarantee the mathematically optimal global and local alignment!'
Exact dynamic programming matrices scale quadratically with sequence lengths, prompting the invention of heuristic seed-and-extend algorithms like FASTA and BLAST.
Phylogenetic Outgroup Disaster
Constructing Maximum Likelihood tree of Primates.
An outgroup is required to polarize character states and root phylogenetic trees to establish the temporal direction of evolutionary divergence.
Unix Command Line Bio Wizardry
Non-bioinformatician: 'How do I extract gene names from this 40 GB GTF file?'
Unix pipes and coreutils (grep, awk, sed, cut, sort) stream large genomic flat files line-by-line with minimal RAM overhead, outpacing Excel by orders of magnitude.
Excel Renaming Human Genes
Genomics dataset opened in Microsoft Excel.
Excel auto-converting gene symbols like SEPT2 (Septin 2) and MARCH1 into calendar dates corrupted >20% of published genomics spreadsheets, forcing HUGO to rename genes.
Bonferroni Correction Massacre
Student running RNA-Seq differential expression testing 25,000 genes with standard p < 0.05:
Benjamini-Hochberg FDR / Bonferroni: '1,249 of those are false positives from multiple testing. Sit down.' 📉
Testing thousands of genes simultaneously inflates Type I family-wise error rates, requiring False Discovery Rate (FDR) corrections like Benjamini-Hochberg or q-values.
RPKM vs FPKM vs TPM Turf War
How do you start a fistfight at a bioinformatics conference?
TPM (Transcripts Per Million) normalizes for sequencing depth after accounting for gene length, ensuring the sum of all TPMs is identical across samples unlike RPKM/FPKM.
Hypothetical Protein Ubiquity
What is the most common protein in GenBank and UniProt?
Automated gene prediction software flags open reading frames (ORFs) that lack characterized homologs, classifying them as hypothetical or uncharacterized proteins.
Git Commit Versioning Reality
Project directory of a PhD student:
Version control systems like Git provide immutable commit hashes and reproducible tracking, preventing the chaotic sprawl of fragmented duplicate files.
K-mer Assembly Velvet Nightmare
Why do de novo genome assemblers hate repetitive transposons?
De Bruijn graph assemblers split short NGS reads into k-mers; repeats longer than read length create bifurcating cycles that fragment assemblies into thousands of contigs.
