It's kind of a "negative" result, but I'll add this one to the mix, as I think it highlights the importance of being careful when doing large-scale computational analyses with complex genomics data:

• Major data analysis errors invalidate cancer microbiome findings

Here are a couple of possibilities that immediately came to my mind:

• A draft human pangenome reference
• Protein-to-genome alignment with miniprot

In addition to what others have already posted:

The long-read assembly papers have been pretty influential (or rather, will be very influential in the years to come):

• "The complete sequence of a human Y chromosome"
• "Telomere-to-telomere assembly of diploid chromosomes with Verkko"
• "The complete sequence of a human genome"
• "Resolution of structural variation in diverse mouse genomes reveals chromatin remodeling due to transposable elements"
  • Many others...

For non-consortium papers that I'd say are influential (in part, basing them on social media response):

There's "The specious art of single-cell genomics" (Plos comp bio) by my colleague, which has ignited some discussion+debates+considerations about t-SNE/UMAPs.

There's "Major data analysis errors invalidate cancer microbiome findings" (Mbio), which has performed some important re-analysis of a major finding and revealed how important it is to normalize correctly and to make absolutely sure your reads are aligning to what you think they are aligning to. Processing genomics data (at both the read-level and quantification-level) is very difficult to get "right" so be extra wary of your own papers and of papers by others when drawing large biological conclusions from one genomics data analysis.

• Edit: Someone else beat me to this as I was typing :)

• https://www.nature.com/articles/s41586-023-06415-8
• https://www.pnas.org/doi/abs/10.1073/pnas.2216697120
• https://www.nature.com/articles/s41467-023-38328-5