TheSequence Radar #674: Transformers in the Genome: How AlphaGenome Reimagines AI-Driven Genomics
A model that could advance the future genomics.
Next Week in The Sequence:
Knowledge: An intro to the world of multi-agent benchmarks.
Engineering: Let’s hack with the Gemini CLI.
Opinion: Why circuits could be the answer to AI interpretability?
Research: AlphaGenome deep dive.
Let’s Go! You can subscribe to The Sequence below:
📝 Editorial: Transformers in the Genome: How AlphaGenome Reimagines AI-Driven Genomics
I have been obsessed with AI in genetics for some time so I couldn’t write about anything else today other than DeepMind’s new model: AlphaGenome!
AlphaGenome merges some of the best -established techiques in AI-driven genomics such as large-scale sequence context with base-pair precision to chart the regulatory genome in a way never before possible. The model’s four-headed architecture digests up to one million contiguous base pairs in a single pass, outputting synchronized predictions for chromatin accessibility, transcription-factor occupancy, RNA expression, splicing, and 3D genome architecture. This unified approach replaces fragmented, single-modality pipelines—each requiring separate models and datasets—with one cohesive model that excels across tasks, streamlining variant effect analysis for researchers.
At its core, AlphaGenome marries convolutional layers, which capture local nucleotide motifs analogous to transcription-factor binding sites, with transformer modules that integrate distal regulatory elements hundreds of kilobases apart. DeepMind’s design eschews downsampling, ensuring every nucleotide contributes to high-resolution inferences. As functional genomics datasets from consortia like ENCODE, GTEx, and 4D Nucleome expand, this backbone stands ready to unveil regulatory grammar buried deep in non-coding DNA.
Traditional genomics models often excel at one signal—SpliceAI for splicing, ChromBPNet for chromatin state—necessitating an ensemble of tools to profile variant consequences fully. AlphaGenome’s simultaneous four-headed predictions eliminate this bottleneck, revealing cross-modal interactions—e.g., how a variant that disrupts a splice site may also alter local chromatin loops—opening novel avenues for mechanistic insight.
In benchmark evaluations spanning 24 sequence-prediction and 26 variant-effect tasks, AlphaGenome matches or surpasses specialized baselines in over 90% of cases. It outperforms SpliceAI, ChromBPNet, and other state-of-the-art models by significant margins, all while completing variant-effect scans in under a second—transforming in silico hypothesis testing from minutes or hours to real-time speed.
The genomics market in 2025 stands at an inflection point: cloud-based sequencing costs have halved over five years, single-cell and long-read technologies have become routine, and multi-omic datasets proliferate. Yet, analytical bottlenecks limit the translation of raw data into actionable insights. AlphaGenome arrives precisely when biotechnology and pharmaceutical companies demand scalable, AI-driven interpretation to bridge the gap from variant discovery to biological understanding. Its ability to standardize and accelerate regulatory variant annotation is poised to catalyze next-generation diagnostic tools, precision therapeutics, and synthetic biology platforms, redefining competitive advantage in a data-saturated market.
DeepMind’s preview API grants non-commercial researchers early access to AlphaGenome, democratizing large-scale regulatory predictions. From pinpointing causal non-coding mutations in disease cohorts to engineering synthetic enhancers with bespoke cell-type specificity, this open sandbox invites collaborative breakthroughs across academia and industry.
If AlphaFold decoded protein structures, AlphaGenome now deciphers the regulatory code—the “dark matter” governing gene expression. As single-cell, long-read, and cross-species datasets proliferate, the model’s extensible architecture promises seamless integration of new modalities. The future of genomics is computational, and AlphaGenome lights the path forward: an intellectual and technological leap toward understanding—and ultimately rewriting—the language of life.
🔎 AI Research
AlphaGenome: advancing regulatory variant effect prediction with a unified DNA sequence model
AI Lab: Google DeepMind
Summary: AlphaGenome is a deep learning–based sequence-to-function model that ingests one megabase of DNA sequence and predicts thousands of functional genomic tracks—including gene expression, transcription initiation, chromatin accessibility, histone modifications, transcription factor binding, chromatin contact maps, and splicing—at single-base-pair resolution. Trained on both human and mouse experimental data, it unifies long-range sequence context with high prediction resolution, outperforming prior methods and enabling comprehensive in silico characterization of regulatory variant effects.
Confidential Inference Systems: Design Principles and Security Risks
AI Lab: Pattern Labs / Anthropic
Summary: This whitepaper defines the architecture of a “confidential inference system” that leverages hardware-based Trusted Execution Environments (TEEs) to protect both user data (model inputs/outputs) and model assets (weights and architecture) during AI inference workloads. It further details reference designs for secure model provisioning, enclave build environments, service provider guarantees, and a comprehensive threat model to mitigate systemic and implementation-introduced risks.
USAD: Universal Speech and Audio Representation via Distillation
AI Lab: MIT CSAIL
Summary: USAD distills knowledge from multiple domain-specific self-supervised audio models into a single student network capable of representing speech, music, and environmental sounds. By training on a diverse multimedia corpus with layer-to-layer distillation, it achieves near state-of-the-art performance across frame-level speech tasks, audio tagging, and sound classification.
UniVLA: Unified Vision-Language-Action Model
AI Lab: CASIA / BAAI / Tsinghua University / HKISI
Summary: UniVLA reformulates vision, language, and robotic actions into shared discrete tokens and learns them jointly in an autoregressive transformer, eliminating separate modality encoders or mapping modules. This unified approach, trained on large-scale video datasets, sets new benchmarks on multi-stage robot manipulation tasks like CALVIN and LIBERO.
ProtoReasoning: Prototypes as the Foundation for Generalizable Reasoning in LLMs
AI Lab: ByteDance Seed / Shanghai Jiao Tong University
Summary: ProtoReasoning introduces “reasoning prototypes”—abstract Prolog and PDDL templates—that capture common logical patterns across diverse tasks and guides LLMs to translate problems into these prototypes. Automated prototype construction and verification via interpreters boosts model generalization and reasoning performance on out-of-distribution benchmarks.
Reinforcement Learning Teachers of Test-Time Scaling
AI Lab: Sakana AI
Summary: This work trains compact “Reinforcement-Learned Teachers” that ingest both questions and ground-truth solutions to learn dense rewards aligned with student performance, departing from sparse-reward paradigms. A 7B-parameter teacher model surpasses much larger reasoning models on competition-level math and science benchmarks and transfers zero-shot to novel tasks.
🤖 AI Tech Releases
Gemma 3n
Google released a full version of Gemma 3n, its mobile optimized model.
Gemini CLI
Google open sourced Gemini CLI, a coding terminal agent powered by Gemini.
Manus Browser
Manus released an agentic browser.
Qwen-VLo
Alibaba open sourced Qwen-VLo, an image understanding and generation model.
🛠 AI in Production
Project Vend
Anthropic showcased Project Vend, a system that allows Claude to run a small shop.
Ray at Pinterest
Pinterest shares how they scale end-to-end ML pipelines with Ray.
📡AI Radar
Meta has successfully recruited Lucas Beyer, Alexander Kolesnikov, and Xiaohua Zhai—founders of OpenAI’s Zurich office—to its new “superintelligence team” in what’s being called Zuckerberg’s latest recruiting victory.
Anthropic launched its Economic Futures Program to support research and policy development.
Uber is in talks with Travis Kalanick to find autonomous car company Pony.AI.
Prediction market Kalshi closed a $185 million Series B round led by Paradigm at a $2 billion post-money valuation, even as rival Polymarket reportedly eyes a $200 million raise.
Data management firm Rubrik announced an agreement to acquire Predibase to speed enterprise adoption of agentic AI—from pilot deployments to production at scale.
Battery startup Nascent Materials emerged from stealth after raising $2.3 million to commercialize an energy-efficient process that produces uniformly sized LFP cathode particles for higher-density, lower-cost batteries.
E-commerce veteran Julie Bornstein’s startup Daydream is launching an AI-powered chatbot tailored for fashion shopping following its $50 million seed round.
AI medical scribe Abridge secured $300 million in a Series E to double its valuation to $5.3 billion, led by Andreessen Horowitz with participation from Khosla Ventures.
Voice-to-text app Wispr Flow raised $30 million in Series A funding from Menlo Ventures (with NEA, 8VC, and angel investors) to scale its AI-powered dictation software across Mac, Windows, and iOS.
Andy Konwinski, co-founder of Databricks and Perplexity, pledged $100 million of his own funds via the Laude Institute to back AI research grants and the new AI Systems Lab at UC Berkeley.
Legal-focused AI startup Harvey AI raised $300 million in Series E funding at a $5 billion valuation—just four months after its prior $3 billion round—to expand its automation tools beyond law into professional services.
European challenger bank Finom closed a €115 million Series C led by AVP, bringing its total funding to ~$346 million as it ramps up AI-enabled accounting and targets 1 million SMB customers by 2026.
Creatio unveiled its 8.3 “Twin” release, embedding a unified conversational interface and new role-based AI agents for CRM and workflow automation along with AI-powered no-code development tools at no extra cost.
Nvidia shares have surged back to a record week, positioning the company within striking distance of a $4 trillion market capitalization as demand for its AI chips continues to accelerate.
Audos, the AI-powered startup studio, aims to democratize entrepreneurship by using AI agents and social-media distribution to launch up to 100,000 companies annually without taking equity.