Arun Das

A complete diploid human genome benchmark for personalized genomics Human genome resequencing typically involves mapping reads to a reference genome to call variants; however, this approach suffers from both technical and reference biases, leaving many duplicated and ...

Delighted to finally announce a preprint describing the Q100 project! “A complete diploid human genome benchmark for personalized genomics” For which we finished HG002 to near-perfect accuracy: www.biorxiv.org/content/10.1... 🧵[1/14]

I just cannot see academic institutions, medical facilities and so many other employers stumping up $100K.

It’s disheartening, both to see this happen and to see how few people who rely on and employ individuals on H-1B visas are willing to speak up and address this. Your silence is deafening.

Make no mistake, this remains catastrophic.

It makes it near impossible for people to stay and work in the US, no matter how qualified they are, unless they work for a handful of extremely wealthy companies.

For so many of us, it’s time to make other plans.

Small victories, but this doesn’t seem to apply to those currently on an H-1B visa.

Wish it was made clear in the initial “proclamation”, before we spent the entire day panicking while trying to figure out a way to get a friend back to the US before midnight.

This is catastrophic.

So, so many people I know and love are going to find it impossible to stay and work in the US, and it makes it almost impossible for people like me to stay and work here in the longer term, no matter how qualified we are.

New blog post – A quick look at Roche's SBX
lh3.github.io/2025/09/11/a...

The concessions made by Brown endanger so many members of our community on campus, limit the access to higher education for individuals from underrepresented backgrounds, and undermine the serious conversations about major issues on campus.

Just so disappointing to see them go down without a fight.

Brown University Makes a Deal With the White House to Restore Funding

Brown have agreed to do all this without any sort of legal challenge, and have agreed to these terms without consulting their staff, students or alumni.

Here's a link to the article that should bypass the paywall, if you wanted to read it for yourself. www.nytimes.com/2025/07/30/u...

In short, all funding is restored and active cases are dismissed in exchange for a $50M commitment to state work force development, new compliance with the administration's discriminatory policies on transgender individuals, a slew of "anti-DEI" admissions policies. No admission of any wrongdoing.

Extremely disappointed in my alma mater, who have chosen to fold without a fight and endanger the most vulnerable members of our community instead of standing up for them.

Spineless and shameful.

Getting into computational biology this summer? 🏖️ 📖 Check out “Beyond the Human Genome Project: The Age of Complete Human Genome Sequences and Pangenome References” by @arun-das.bsky.social‬, @mikeschatz.bsky.social, and more for a great introduction to the field:

Partitioned Multi-MUM finding for scalable pangenomics Pangenome collections are growing to hundreds of high-quality genomes. This necessitates scalable methods for constructing pangenome alignments that can incorporate newly-sequenced assemblies. We prev...

Excited to share a new update to Mumemto, scaling MUM and conserved element finding to any size pangenome! Preprint out now w/ @benlangmead.bsky.social.
Mumemto scales to the new HPRC v2 release and beyond, and can merge in future assemblies without any recomputation! 1/n

Easily the most important thing happening next week.

Come and watch my friend Sara defend her PhD!

Thank you! It was awesome to talk to you too, and to learn about all the cool data and insights from your project!

Really cool work from @arun-das.bsky.social on recovering sequence from unmapped reads (even with T2T reference or HPRC pangenomes!). Can recover a decent amount of sequence per individual using these approaches. Check it out!

@arun-das.bsky.social's thesis research demonstrates that short-read mapping-based approaches, even using complete linear (T2T-CHM13) and pangenome (HPRC) references, miss a lot of variation that can be recovered from unmapped reads.

GitHub - arun96/SouthAsianGenomeDiversity: Repository containing all links and materials relevant to our work on investigating the diversity in and the representation of South Asians in genomic datase... Repository containing all links and materials relevant to our work on investigating the diversity in and the representation of South Asians in genomic datasets. - GitHub - arun96/SouthAsianGenomeD...

The homepage for this work is here: github.com/arun96/South...

This analysis can be replicated on any population of your choosing, and the WDL scripts used to run the various stages (as well as any other analysis details) can all be found on that page and in our pre-print.

Comparison of the significant loci associated with HDL cholesterol in the South Asian set in the UK Biobank to that of the East London Genes and Health cohort, alongside our placed contigs against GRCh38.

Finally, we compare our placed contigs to loci associated with biomarker traits in the UK Biobank and East London Genes & Health Dataset, and find a number of positions where a placed contig is close to a significant locus.

Plot showing the distribution of the most aligned-to contigs in our RNA-seq contigs, against the length of the contigs and colored by their population. The contigs vary widely in terms of their RNA-seq alignment density and their lengths.

We are also able to align existing RNA-Seq data from 140 SAS individuals from MAGE directly to these contigs, allowing us to identify 200 contigs with a high density of RNA-Seq alignments.

BLAST shows that these contigs are highly similar to non-reference human and primate sequences.

Comparison of our insertions to Manta. We find an order of magnitude more large insertions than this tool, and comparable amounts to existing large insertion callers.

We show that the majority of the placements we make are missed by traditional insertion calling tools, but in line with specific large non-reference sequence detection ones.

For the unplaced contigs, BLAST shows that the majority have high similarity to non-reference human and primate sequences.

Visualization of placements throughout CHM13. We place contigs all over the genome.

Plot of the number of unique gene intersections per chromosome. We see the number of intersections are largely correlated with chromosome size.

Key genes we find intersected with contigs.

We are able to place ~20K contigs against CHM13 through a combination of alignment, mate pair read information and LD.

We find >8,000 instances of a placed contig intersecting one of 106 protein coding genes, and >6,000 placements within 1 Kb of a known GWAS site.

Plot of the fraction of assembled contigs from each of 21 SAS individuals that are validated by their long read assembly, ~85% of the contigs per individual are validated.

Plot of the amount of assembled sequence per individual across two linear and two pangenome references. Massive reductions in sequence are seen as we go from GRCh8 to CHM13, but the HPRC pangenomes offer only a slight improvement after that.

We validate >80% of these contigs in a subset of 21 SAS individuals using auxiliary long read data.

We repeat the linear pipeline with the HPRC v1 draft pangenomes, and see further improvements in alignment but only small reductions in the amount of assembled sequence.

Comparison of alignment rate against GRCh38 and CHM13. We see a 0.5-1% improvement against CHM13.

Histogram of amount of assembled sequence across 640 SAS individuals. We assemble on average 550-600 Kb per individual from unmapped reads.

Despite improvements in alignment compared to GRCh38, we assemble ~600 Kb of sequence in >1 Kb contigs per individual from unmapped reads against T2T-CHM13.

Across the whole set, we assemble 410 Mb of sequence in 199K contigs (which collapses down to 50 Mb when accounting for shared sequence).

Our analysis pipeline, which consists of 1) Aligning existing reads against reference genomes, 2) assembling unaligned or poorly aligned reads, 3) placing the large assembled contigs back into the reference, 4) calling variants and novel sequence, and 5) evaluating the functional impact of this variation.

Source of our data. 601 individuals come from 5 1KGP populations, 39 come from 19 SGDP populations.

To do this, we align existing short read data from 640 South Asian (SAS) individuals from 1KGP and SGDP against linear & pangenome references, and assemble the unmapped reads into large contigs.

We then attempt to analyze the functional impact of these sequences.

Comparison of the fraction of individuals in the GWAS catalog of different ancestries to the breakdown of the global population in terms of those ancestries. South Asians accounted for 2% of the GWAS catalog in 2019, but for >25% of the global population.

South Asians are severely underrepresented in genomics, and this lack of representation makes it difficult to catalog and understand the variation present in these communities.

Our goal was to investigate the variation present in these populations that is missing in widely used reference genomes.

Our pre-print on investigating variation in South Asian genomes is now out!

Thank you to @mikeschatz.bsky.social, @rajivmccoy.bsky.social and @aabiddanda.bsky.social for all their work on this.

🧵 A thread on the key results and takeaways from our work:

Happening at 2pm in Biondi - come and check out Vikram’s great work!

Thank you to everyone who attended, and thank you to everyone who got me to this point - I appreciate you all.

I'm on the job market this summer, so please send any interesting opportunities my way 😃

I'll also be on the job market this summer, so please reach out if you're interested!

You can find out more about me at these links:
LinkedIn: www.linkedin.com/in/arun96/
Personal Website: arundas.org

If any of that interests you, or you would like the link to my defense, please shoot me a DM here or contact me through any one of these ways: arundas.org/Contact.html .