iiinova — where your digital self keeps evolving

Dani León-Periñán @danilexn.bsky.social · 28/09/2026

Especially - given the recent Anthropic & Open-AI fiascos -, when you let AI come up with the ideas behind your manuscript, you’re potentially stealing somebody else’s work :)

000

Dani León-Periñán @danilexn.bsky.social · 28/09/2026

Reading fully-AI-generated manuscripts is extremely frustrating.

130

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

Under the hood, Malva is a k-mer method. In the last years, this field has seen lots of *great* ideas and applications, and we are happy to contribute a small part to this. You can find out all details in the paper!

030

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

@nukappa.bsky.social, Nikolaus Rajewsky, and I have been building this for a few years at @mdc-berlin.bsky.social. We are very excited about it and hope it becomes useful to lots of you. Please share your feedback so we can keep improving the platform for everyone! 🤗

130

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

The Malva Index stores every read tied to its cell, so you can transpose it and cluster cells by sequence alone. No reference genome needed, which means this generalizes to organisms with no assembly at all.

Conceptual view of the single-cell information funnel. Biological (‘physical’) cells are converted to ‘digital’ cells via sequencing, which can be represented either as cell-by-gene or cell-by-sequence matrices. These representations enable reconstruction of latent spaces underlying cell types and trajectories. On the right, Spatial clustering of a head and neck tumour dataset using gene count-based (left) versus sequence-based (Malva, right) embeddings. KS1 and KS2 are two k-mer–specific clusters

110

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

We think efficient sequence search across large single-cell data is a key step toward understanding what RNAs actually do. When you interrogate where an event (a splice junction, a 3'UTR, a TE) is present across hundreds of cell types and conditions, you find patterns that lead to function.

Malva analysis explains 3′ UTR usage patterns in its spatial context—that is, 3′ UTR length
regulation of Add2 during spatiotemporal mouse embryogenesis

120

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

So we deployed a large server at @mdc-berlin.bsky.social to crawl public repositories and index every read, using an algorithm we crafted for cellular resolution. In a few hours, it ingested >500 TB of SC data into one index across hundreds of tissues, cell types, diseases... that grows daily.

Public sequence archives (NCBI GEO, SRA, ENA, CNGBdb and the Human Cell Atlas (HCA) data portal) are continuously crawled for single-cell RNA-seq (scRNA-seq), single-nucleus RNA-seq (snRNA-seq) and spatial RNA-seq datasets, which are indexed and integrated into Malva Index. On the right, cumulative expansion of the human Malva Index (v2025.04a), with sequencing volume stratified by organ system and technology type.

130

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

Malva is a new member of your single-cell analysis toolbox. We noticed that portals are useful, but precomputed gene counts make isoforms, mutations, and pathogen sequences invisible.

A researcher tries to search “Find cells expressing sequence XYZ across all Human Cell Atlas” but current tools (e.g., single-cell databases and sequence search algorithms) do not resolve the query (no sequence search or no single-cell resolution, respectively). With Malva, researchers can answer those sorts of questions, since it is based on indexing of sequences at single-cell resolution

130

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

Malva is live today for academic users, so you can search for any sequence across millions of cells and see where it sits relative to everything else. You can use Malva from the browser, through the API, or via AI agents. The platform is here (you just need an ORCID): malva.mdc-berlin.de

malva.mdc-berlin.de

Home — Malva

120

Dani León-Periñán @danilexn.bsky.social · 26/08/2026

Malva out in Nature! 🚀 RNA sequences make cells different from each other. Single-cell data captures much of this, but most stays in the dark because we reduce it to gene counts. We introduce Malva to make the world's single-cell data searchable. What took days is now one query in seconds.

1207

Dani León-Periñán @danilexn.bsky.social · 27/02/2026

This is such an exciting project, join us!

010

Reposted by Dani León-Periñán

Nikos Karaiskos @nukappa.bsky.social · 27/02/2026

🚀 We're looking for people to join Malva! 🧬 @danilexn.bsky.social Malva is the first search engine for single-cell genomics: we process massive amounts of sequencing data to enable ultrafast analysis of millions of cells. (1/4)

154

Reposted by Dani León-Periñán

Nature Methods @natmethods.nature.com · 14/01/2026

SpatialZ generates virtual single-cell transcriptomics slices between experimentally measured sections, enabling accurate and efficient building of 3D cell atlases of different tissues. www.nature.com/articles/s41...

0105

Reposted by Dani León-Periñán

Damian Carrington @dpcarrington.bsky.social · 13/01/2026

‘A bombshell’: doubt cast on discoveries of #microplastics throughout human body, from brain to blood, arteries to testes Exclusive: Some scientists say many detections are most likely error, with one high-profile study called a ‘joke’ Story by me www.theguardian.com/environment/...

theguardian.com

‘A bombshell’: doubt cast on discovery of microplastics throughout human body

Exclusive: Some scientists say many detections are most likely error, with one high-profile study called a ‘joke’

1421670

Dani León-Periñán @danilexn.bsky.social · 03/01/2026

nope, US rn is worse than Nazi Germany. you have a fascist government AND nuclear weapons.

000

Reposted by Dani León-Periñán

Martin Pacesa @martinpacesa.bsky.social · 26/12/2025

Repeat after me: GENE 👏 EXPRESSION 👏 PROFILING 👏 DOES 👏 NOT 👏 A 👏 VIRTUAL 👏 CELL 👏 MAKE 👏

47916

Reposted by Dani León-Periñán

Kris Willis @kawillis.bsky.social · 18/12/2025

Today is a big day. After many years of work, I’m excited to finally share a paper describing a novel approach to identifying potential breakthroughs in biomedical research, up to twelve years before the breakthrough itself occurs. 1/15

1118

Reposted by Dani León-Periñán

Bioinformatics Advances @bioinfoadv.bsky.social · 05/12/2025

🧮 Just out in Bioinformatics Advances: “FroM Superstring to Indexing: A space-efficient index for unconstrained k-mer sets using the Masked Burrows-Wheeler Transform (MBWT)” Full article available: doi.org/10.1093/bioadv/vbaf290 Authors include: @pavelvesely.bsky.social, @brinda.eu

1113

Dani León-Periñán @danilexn.bsky.social · 06/11/2025

Needless to say that the bioinformatics analyses presented in the manuscript show all the common pitfalls/cliches. Worst of all: no judgement of methods. Just plain belief that outputs of code interpreters are ground truth, and any contradictions are “conceptual” rather than “methodological”.

030

Dani León-Periñán @danilexn.bsky.social · 06/11/2025

… and developing those tools and standards requires time, effort, spent by bright people and not by sloppy AI agents. You cannot claim to have an “AI Scientist for Autonomous Discovery”, when your scRNA analysis relies on tools that are created and maintained by humans. That’s not autonomous.

120

Dani León-Periñán @danilexn.bsky.social · 06/11/2025

I see the value of agent-based AI. We work on developing stuff in that direction. But claims like “Here we present Kosmos, an AI scientist that automates data-driven discovery” are just FALSE and malicious. Discovery in bioinformatics requires an ecosystem of tools and good practices…

110

Dani León-Periñán @danilexn.bsky.social · 06/11/2025

From the manuscript: “collectively executing an average of 42,000 lines of code and reading 1,500 papers per run” Looking at those 42,000 lines of code, most are bioinformatics 101, stolen from tutorials/books. Many are just repeated across notebooks.

120

Dani León-Periñán @danilexn.bsky.social · 06/11/2025

Also, the whole concept of “data-driven discovery” is super funny. In order to get good data, there has to be very good, careful experimental design. And experimental work. And most time, lots of new methods development.

110

Dani León-Periñán @danilexn.bsky.social · 06/11/2025

This is pure bullshit. You don’t need to run 200 LLM agents for >8 hours to do DE, GSEA and generate a couple of heatmaps. And if you do, just spend 2 weeks learning bioinformatics and you won’t need to give money to a company selling snake oil. arxiv.org/pdf/2511.02824

arxiv.org

251

Reposted by Dani León-Periñán

Champi @jpascualanaya.bsky.social · 30/10/2025

🚨 We're hiring! 🚨 Our lab at the University of Málaga, Spain🇪🇸 is looking for a bioinformatician researcher for a postdoc position (1+1 years). Check the ad, and if you're interested, get in touch! 📅 Starting date early 2026, with some flexibility. 🙏Please, RT!

02738

Dani León-Periñán @danilexn.bsky.social · 29/10/2025

Esto acabará cuando a alguien que haya perdido a un ser querido se le vaya la cabeza, y extermine a la rata.

010

Reposted by Dani León-Periñán

Sam Gershman @gershbrain.bsky.social · 15/10/2025

When I was an undergrad, I saw a talk by Oliver Sacks about music. During the Q&A, someone asked Sacks if he thought science would reveal the deepest mysteries of art. Sacks (a rare humanist among scientists) said that he doubted it, at which Eric Kandel (the host) leaped up and grabbed the mic.

1617023

Reposted by Dani León-Periñán

Dr. Jean Fan @jef.works · 13/10/2025

What is RNA velocity in situ? ChatGPT kept getting it wrong... So I made this video explaining how RNA velocity in situ infers gene expression dynamics by distinguishing nuclear vs. cytoplasmic expression in spatial transcriptomics data: youtu.be/z9Oxf1hvum0 #AcademicSky

youtu.be

RNA velocity in situ infers gene expression dynamics using spatial transcriptomics data

YouTube video by Prof. Jean Fan

2268

Reposted by Dani León-Periñán

Robert Aboukhalil @robert.bio · 09/10/2025

Just published an interactive article about a magical algorithm known as the Burrows-Wheeler Transform, which powers sequence alignment tools like bowtie and bwa: sandbox.bio/concepts/bwt It's also notoriously unintuitive so I'm hoping this article helps you build that intuition.

39829

Reposted by Dani León-Periñán

Camille Marchet ⚡ @camillemrcht.bsky.social · 09/10/2025

Ca n'est pas si souvent, un article publié dans Nature met ma communauté à l'honneur (la bioinformatique des séquences). Je vous raconte ? www.nature.com/articles/d41...

nature.com

‘Google for DNA’ brings order to biology’s big data

MetaGraph compresses vast data archives into a search engine for scientists, opening up new frontiers of biological discovery.

13114

Reposted by Dani León-Periñán

Sheila Teves @sheilateves.bsky.social · 07/10/2025

How does a “universal” transcription factor evolve to do species-specific jobs? Our new preprint reveals how divergence in TBP’s domains shapes transcriptional specialization across eukaryotes. Read it here 👉 www.biorxiv.org/content/10.1... #MolecularEvolution #Transcription

biorxiv.org

Molecular determinants underlying functional divergence of TBP homologs

The TATA-box binding protein (TBP) is a highly conserved basal transcription factor and a core component of the pre-initiation complex (PIC) for all three eukaryotic RNA polymerases (RNA Pols). Despit...

24216

Reposted by Dani León-Periñán

Dr. Damien P. Williams, dread portent down from a mountain cave @wolven.blacksky.app · 05/10/2025

Preliminary results show that the current framework of "AI" makes ppl less likely to help or seek help from other humans, or to seek to soothe conflict, and that people actively prefer that framework to any others, literally serving to make them more dependent on it.

arxiv.org

Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence

Both the general public and academic communities have raised concerns about sycophancy, the phenomenon of artificial intelligence (AI) excessively agreeing with or flattering users. Yet, beyond isolat...

15455215

Dani León-Periñán @danilexn.bsky.social · 01/10/2025

Yes, reading it rn! Super cool, congrats!!!

010

Dani León-Periñán @danilexn.bsky.social · 01/10/2025

Tweet only for Spanish native speakers

000

Reposted by Dani León-Periñán

Merriam-Webster @merriam-webster.com · 01/10/2025

We are thrilled to announce that our NEW Large Language Model will be released on 11.18.25.

647249378241

Reposted by Dani León-Periñán

Paul Medvedev @pashadag.bsky.social · 12/06/2025

1/4 Hash functions in genomic sequence analysis (tinyurl.com/4kk9ccmt) : a new survey written together with Ke Chen, Xiang Li, Qian Shi, and Mingfu Shao. Before submitting it, we are posting it online to get feedback from the community.

tinyurl.com

Dropbox

12816

Reposted by Dani León-Periñán

EL PAÍS @elpais.com · 26/09/2025

🔴 ÚLTIMA HORA | La mayoría de los países abandonan la Asamblea General de la ONU a la llegada de Netanyahu social.elpais.com/njzv3

741657633

Dani León-Periñán @danilexn.bsky.social · 26/09/2025

Mr. Big Ears said at the UN today, “Thanks to the efforts of our intelligence service, this message is being streamed to every Gazan’s phone.”

000

Reposted by Dani León-Periñán

Yun S. Song @yun-s-song.bsky.social · 22/09/2025

We are excited to share GPN-Star, a cost-effective, biologically grounded genomic language modeling framework that achieves state-of-the-art performance across a wide range of variant effect prediction tasks relevant to human genetics. www.biorxiv.org/content/10.1... (1/n)

417690

Dani León-Periñán @danilexn.bsky.social · 22/09/2025

IMSLP (imslp.org; the biggest public library for classical music) generated >4 TB of data over the last 20 years. Open-ST (rajewsky-lab.github.io/openst) can generate that from a 1x3x4 mm piece of human tissue in two days.

imslp.org

IMSLP: Free Sheet Music PDF Download

000

Reposted by Dani León-Periñán

🧬Jacob L Steenwyk @jlsteenwyk.bsky.social · 13/09/2025

How come different people giving the same prompt to ChatGPT get different answers? A new article from ThinkingMachines explains why LLMs are nondeterministic, and it largely comes down to how parallelization and precision are handled. 🧵

142

Reposted by Dani León-Periñán

Proceedings of the National Academy of Sciences @pnas.org · 28/08/2025

A machine learning model can sort people with myalgic encephalomyelitis/chronic fatigue syndrome from healthy controls with 77% accuracy by sequencing the cell-free RNA in their plasma, suggesting a test for the condition may be on the horizon. In PNAS: www.pnas.org/doi/10.1073/...

Biological pathway analysis of myalgic encephalomyelitis/chronic fatigue syndrome. Heatmap of activated (blue) and inhibited (red) pathway marker genes for the selected pathways (n = 69 genes).

0103

Reposted by Dani León-Periñán

Eric Topol @erictopol.bsky.social · 28/08/2025

This is how generative A.I. could be used to design far better, more potent mRNAs (linear and circular) for the future for genome editing, engineered T cells, cancer therapy, vaccines www.science.org/doi/10.1126/... @science.org

science.org

Deep generative models design mRNA sequences with enhanced translational capacity and stability

Despite the success of mRNA COVID-19 vaccines, extending this modality to more diseases necessitates substantial enhancements. We present GEMORNA, a generative RNA model that utilizes Transformer arch...

314039