Basecamp Research, a cutting-edge AI lab focused on biological design, has unveiled the Trillion Gene Atlas—an ambitious scientific program aimed at generating and modeling genomic data at an unprecedented trillion-gene scale. Developed in collaboration with Anthropic, Ultima Genomics, and PacBio, and supported by NVIDIA’s AI infrastructure, the initiative seeks to increase known genetic diversity by 100 times. It plans to gather genomic information from over 100 million species across thousands of global locations.

This effort builds on Basecamp Research’s expanding network of biodiversity partners worldwide. The long-term vision is to create a vast and diverse dataset that allows AI systems to learn from evolution and enable the on-demand design of new medicines.

Speaking at SXSW in Austin, Co-founder and CEO Glen Gowers noted that current biological AI models rely on a limited representation of Earth’s biodiversity. He explained that the Trillion Gene Atlas will dramatically expand the genetic landscape available for analysis, introducing a new era of programmable therapeutic design powered by large-scale data.

Comparable in scope to the Human Genome Project, the initiative was introduced during SXSW’s Health Track and at the NVIDIA GTC conference in San Jose.

Tackling the Biological Data Gap
Despite rapid growth in model size and computational capabilities, progress in AI-driven drug discovery has been constrained by limited data diversity. Most existing sequence-based models depend heavily on a small set of public databases, with a large portion trained on fewer than 250 million genetic sequences.

To address this, Basecamp Research introduced its EDEN foundation models earlier this year. These models are trained entirely on BaseData™, a proprietary genomic dataset that exceeds the size of all public repositories combined. By incorporating over 10 billion previously unknown genes from one million newly identified species, EDEN has revealed new scaling principles for AI in biology.

This expansion has enabled EDEN to move beyond prediction, allowing it to design therapeutics directly from disease prompts. In laboratory tests, the model demonstrated zero-shot functionality in human T-cells without relying on clinical or human-derived data. It has also produced promising results across multiple advanced applications, including AI-driven gene insertion and the creation of targeted antimicrobial peptides with high success rates.

The Trillion Gene Atlas builds on this foundation by significantly increasing both the scale and contextual richness of genomic data available for AI training.

Expanding a Global Biodiversity Network
Over the past six years, Basecamp Research has established a network of scientific collaborators spanning 31 countries. This has enabled the development of a scalable genomics pipeline designed specifically for AI applications. Using innovative regulatory frameworks and off-grid DNA sequencing technologies, the company is able to collect high-quality genetic data from remote ecosystems often inaccessible to traditional labs.

These partnerships emphasize knowledge sharing, local capacity building, and fair access and benefit-sharing agreements aligned with emerging global standards. As part of the Atlas initiative, new collaborations have been announced in Chile and Argentina, along with expanded research efforts in Antarctica.

Advancing Sequencing and Computing Capabilities
The Trillion Gene Atlas is made possible by breakthroughs in high-throughput sequencing and accelerated computing. Partnerships with Ultima Genomics and PacBio enable large-scale sequencing, including highly accurate long-read data that preserves detailed genomic context.
Ultima’s latest sequencing platform, the UG200 Series, is designed for industrial-scale genome and multi-omics sequencing at lower costs, making projects like the Atlas feasible. Meanwhile, PacBio’s HiFi sequencing technology provides precise, information-rich data critical for training advanced biological AI systems.

NVIDIA’s computing infrastructure will power the processing of massive genomic datasets at the petabase level. By leveraging tools like NVIDIA Parabricks, Basecamp aims to dramatically accelerate metagenomic analysis. Tasks that previously could have taken over two decades are now expected to be completed in under two years through parallel processing, automation, and large-scale model training.

Toward End-to-End AI-Driven Therapeutic Design
Anthropic is contributing to the initiative by integrating its AI system, Claude, with scientific platforms. The goal is to combine Claude’s reasoning capabilities with EDEN’s therapeutic design functions and NVIDIA’s data processing tools to create a seamless workflow—from interpreting complex biological data to generating targeted treatments.

Built on three core pillars—large-scale DNA sequencing, global data partnerships, and advanced computing—the Trillion Gene Atlas represents a major step toward transforming how biological data is used. By expanding evolutionary datasets 100-fold, Basecamp Research aims to accelerate drug discovery, improve precision in therapeutic design, and extend advances in areas such as gene therapy and antimicrobial resistance.

The Leukemia & Lymphoma Society projects that by the end of 2024, approximately 187,740 Americans will be diagnosed with blood cancer, equating to about one new case every three minutes. This statistic highlights the complexity of diagnosing and treating these conditions. Traditional diagnostic methods rely on several laboratory techniques, including karyotyping, fluorescence in situ hybridization, microarrays, gene panels, and PCR testing. These approaches are often intricate, technologically limited, and can yield conflicting results.

Whole-genome sequencing (WGS) offers an unprecedented level of sensitivity and accuracy, enabling the detection of crucial variants and biomarkers for prognosis, risk assessment, and treatment decisions. "Sequencing technology has greatly enhanced clinical outcomes, particularly in cancer risk stratification and diagnosis," says Weida Gong, a bioinformatician at Illumina. "For acute myeloid leukemia (AML), sequencing is vital, as it can alter treatment decisions based on the mutational profile." Certain mutations have a significant impact on disease progression, and if missed, they can lead to poorer survival rates.

Bone marrow and blood cancer patients often undergo multiple tests as part of their standard care. Gong notes that traditional methods like karyotyping or cytogenetics focus on detecting larger chromosomal abnormalities, missing smaller mutations that could be crucial for diagnosis and treatment. These conventional methods may fail to identify mutations that involve small base pair changes, which can be key in understanding the disease.

At the American Society of Human Genetics (ASHG) annual meeting in Denver, Gong presented a study on the performance of WGS in detecting small somatic variants, structural variants (SV), and copy number alterations (CNA) specific to AML. The research, which included 23 clinical samples from Washington University School of Medicine and 30 AML samples from Discovery Life Sciences, as well as over 500 additional samples, demonstrated that WGS provides a more comprehensive and accurate picture of each tumor.

"In AML patients, mutations with low variant allele frequencies (VAF), ranging from 5% to 20%, are critical," Gong explains. "If sequencing is done at lower coverage, these mutations may go undetected." Traditional sequencing methods typically operate at 30× or 40× coverage, whereas Gong's study used 200× coverage, significantly improving the detection of mutations that conventional technologies miss. The study achieved 100% sensitivity, even for hard-to-find insertions and deletions (indels), and detected critical mutations like FLT3-ITD, which is important for AML risk stratification.

In addition, the study found that WGS can detect variants at 95% sensitivity with a 5% VAF at 140× coverage, comparable to the recently FDA-approved TruSight Oncology Comprehensive assay. This ability to detect low-frequency mutations quickly is a key advantage of WGS, as traditional testing often involves multiple steps and takes longer to deliver results.

Illumina has developed a high-coverage WGS method and bioinformatics pipeline specifically for hematological malignancies, enabling faster and more efficient genomic profiling. This approach offers a comprehensive solution for researchers, integrating sequencing, secondary analysis, and interpretation. In early 2025, Illumina will enhance its Connected Insights platform to include automated risk stratification for AML, further improving efficiency and accuracy in clinical research.

While WGS has shown its potential in improving the characterization of hematological cancers, its widespread adoption faces challenges within healthcare systems. Nevertheless, industry leaders recognize the transformative power of WGS, and there is growing interest in its use for blood cancer diagnosis and treatment. "The potential for WGS is limitless," says Gong.

Click here to know more about the Illumina heme WGS interpretation solution.