Forum:What are some major historic shifts in bioinformatic methodologies?

Question

[Jeremy Leipzig]

I'm looking more for major shifts in approaches to analyses that arose from bioinformatic development or borrowed from related fields rather than those borne strictly in response to new wet lab techniques

Some examples I can think of...

• kmer and graph-based methods to genome assembly
• reference-free, pseudoalignments for transcript quantification
• emergence of pipeline frameworks as standard practice
• deep learning methods applied to variant calling
• application of manifold learning/nonlinear dimensionality reduction to visualization (e.g. tSNE)

Answer 1

[GenoMax]

• Development of similarity matrices
• BLAST searching/availability of NCBI Web blast
• Move away from printed GenBank to electronic format
• phred/phrap/consed
• Phylogenetic data analysis methods
• Celera assembler - whole genome shotgun assembly

[Istvan Albert]

an even bigger historic shift will be when we move away from Genbank, what an inefficient and inappropriate format that is

[Mensur Dlakic]

Move away from printed GenBank to electronic format

Can't believe that nobody asked about this. Are all readers of this post that old to know what a printed GenBank is? Next thing you will tell me that everyone here has submitted proteins sequences by e-mail for database searching !!

Answer 2

[lieven.sterck]

• FASTA => BLAST
• read mapping & quantification => pseudo-alignment & quantification (read counts)
• consensus assemblies => haplotype aware assemblies

Answer 3

[4galaxy77]

Figuring out the Burrows-Wheeler transform can be used for data compression / alignment.

Answer 4

[Mensur Dlakic]

• Development of fast profile-based search methods such as PSI-BLAST. The original PSI-BLAST paper has 77502 citations.
• HMMer gave rise to Pfam and many other HMM databases, including its faster and more sensitive profile-profile cousins such as HHblits.
• CD-HIT for protein clustering and MMseqs2 for ultra-fast protein search and clustering of large databases.
• Rosetta.

Answer 5

[poisonAlien]

Maybe not mainstream yet but it will be soon IMO..

1. Pan-genomes and pan-transcriptomes

Answer 6

[dariober]

What about variance-shrinkage methods?

These allowed expression microarrays first and later RNAseq, ChIPseq, and similar datasets to produce reliable results even with few replicates. Not sure if it fits with your requirement of not being in response to a wet lab advancement. However, I wonder if these methods also sort of backfired since they have made us even more stingy in adding more replicates...