@KyleFarh

Head of Illumina AI Lab. Deep Learning. Human Genetics. Perturb-seq.

 San Francisco, CA
Joined May 2025
  • Tweets83
  • Following1.3K
  • Followers470
5Media
Excited to be introducing Illumina's dnaAI model for the opening talk at ASHG 2026! #ASHG26 #ASHG2026
1
3
25
1,159
Excited to announce SpliceAI2: illumina.com/science/genomic… Open model weights: huggingface.co/illumina-ai/S… Source code: github.com/illumina/spliceAI…

Introducing SpliceAI2: The Next Generation of Splicing and Transcript Isoform Prediction

illumina.com
9
19
610
Excited to be participating in the Arc Virtual Cell Challenge! Check out where Illumina's customized PromoterAI_precomputed model is on the leaderboard -- science.org/doi/10.1126/scie… virtualcellchallenge.org/lea…
2
3
1
26
1,981
Causal Models Need Causal Data - Xaira’s X-Cell model (Bo Wang & Ci Chu) youtube.com/watch?v=2AdS-2uu…

🔬Causal Models Need Causal Data - Xaira’s X-Cell model (Bo Wang & Ci Chu)

Can a model predict how a cell responds to a genetic perturbation i...

youtube.com
514
JEPA-DNA paper learns latent embeddings of genomic sequence rather than next token prediction, enabling models to capture global functional context and go beyond local motifs. JEPA-DNA arxiv.org/abs/2602.17162

JEPA-DNA: Grounding Genomic Foundation Models through...

Genomic Foundation Models (GFMs) typically rely on Masked Language Modeling (MLM) or Next-Token Prediction (NTP) to learn the "Laws of Nature". While effective at capturing local syntax, these...

arxiv.org
2
9
70
9,364
Variant age dating with ancestral recombination graphs to improve rare variant interpretation -- biorxiv.org/content/10.64898…

Tracing the evolutionary histories of ultra-rare variants using variational dating of large...

Ultra-rare variants dominate whole-genome sequencing datasets, yet their interpretation is limited by allele frequency, which provides little information at very low counts and is highly sensitive to...

biorxiv.org
3
14
1,632
Classifying pathogenic vs. benign developmental disorder mutations using recurrent somatic mutations -- codons that are mutated in cancer indicate selection, and in a subset of DD genes this is highly informative. biorxiv.org/content/10.64898…

Using somatic data to aid germline clinical variant interpretation in developmental disorders

Accurate interpretation of rare germline variants remains a major challenge in developmental disorders (DD). Somatic mutation data represent a largely untapped source of evidence for germline variant...

biorxiv.org
2
8
1,259
gVAMP paper uses approximate message passing to achieve state-of-the-art results for polygenic risk scores. cell.com/cell-genomics/fullt…

Joint modeling of whole-genome sequencing data for human height via approximate message passing

Depope et al. develop a computationally efficient statistical algorithm for genomics that, when applied to half a million whole-genome sequences, reduces false discoveries and provides high reported...

cell.com
2
6
1,111
CRISPR-based gene editing for personalized therapies is rapidly advancing, emphasizing the need to address the VUS problem so we know what to edit. nature.com/articles/d41586-0…

Personalized CRISPR therapies could soon reach thousands — here’s how

Nature - A fresh approach to trialling treatments for rare genetic diseases could make their production an economically viable prospect at last.

nature.com
4
791
Nice paper that uses CRISPRi perturb-seq to dissect the Alzheimer's locus at TSPAN14. sciencedirect.com/science/ar…
1
1
10
1,196
Now imagine what if we had perturb-seq data for all genes, in a billion cells across dozens of cell types? illumina.com/areas-of-intere…

NGS-based cell atlas for drug discovery and AI models

Genome‑wide genetic perturbation data set from Illumina built to accelerate drug discovery and AI models

illumina.com
229
Glial cells, not-neurons, utilize the most diverse alternative splicing in the adult cortex. biorxiv.org/content/10.1101/…
1
6
929
Our team has a position in the SF bay area for a top computational + experimental scientist for perturb-seq. illumina.wd1.myworkdayjobs.c…
6
28
4,713
Kyle Farh MD PhD - Illumina AI Lab retweeted
Jonas Koeppel@JonasKoeppelJun 3
1/n What fraction of the human genome is essential for cellular viability? Excited to share our preprint that explores this question by combining an unusual CRISPR system, phage promoters, and thousands of deletion launchpads. @sudpinglay @JShendure biorxiv.org/content/10.64898…
4
53
7
173
30,372
Kyle Farh MD PhD - Illumina AI Lab retweeted
Elliot Hershberg
@ElliotHershbergJun 2
I'm pretty excited about AI in Biology. I bet my research career on it. And now I'm betting my investment career on it! But somehow, spending this past week at CSHL's 90th symposium—which was exclusively dedicated to this topic this year—made me even more bullish! One of the most electric conferences I've been to in a while... maybe ever. Back-to-back talks from the top scientists and industry leaders in this field. Even though my job is to cover this domain, it was a visceral reminder of how quickly things are moving. Fantastic job curating and organizing by @pkoo562 and @TonyZador. My plan is to get some notes up on what I saw in the next week or two on CoB! 🧬
4
16
204
14,097
Kyle Farh MD PhD - Illumina AI Lab retweeted
Anshul Kundaje@anshulkundajeJun 4
Academic labs continue to lead on next generation tech dev for powering AI models for biology. The next big leap in models of gene regulation will be when they truly master the long range functional architecture of the genome. 1/
Sudarshan Pinglay@sudpinglayJun 4
How much of the human genome is essential? Two pieces out today from our lab: 1) a method to map essential genomic intervals at gigabase scale, and 2) an argument that it's time to consider synthesizing a minimal human genome. biorxiv.org/content/10.64898… nature.com/articles/d41586-0…
4
23
172
25,515
Kyle Farh MD PhD - Illumina AI Lab retweeted
Ruslan Rust
@rust_ruslanJun 3
A multimodal perturbation atlas of 1,000 pooled CRISPR knockouts in A549 cells, profiled by fluorescence microscopy (39 live, 13 fixed markers), label-free phase imaging of the same live cells, and single-cell RNA sequencing (scRNA-seq) Totaling ~57 million single-cell profiles Preprint: biorxiv.org/content/10.64898…
2
22
1
83
8,252
New study using PromoterAI to show that disruption of key transcription factor motifs reduces gene expression and further strengthens experimental results (Figure 7). biorxiv.org/content/10.64898…
1
5
1,057
New from Lehner lab: OpenSplice — ~600k variants across 608 exons. Largest splicing saturation mutagenesis dataset by far. SpliceAI still the top performer among 4 modern models tested (incl. Pangolin, SpliceTransformer, AlphaGenome). biorxiv.org/content/10.64898…
4
461
Load more