Nicholas K. Wang

Nicholas K. Wang

Auto-generated profile, based on publicly available information

Nicholas K. Wang

Department of Human Genetics, University of California

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
4 March 2026 in Nature
Verified
Journal Article
Open access
PDF file
GBGaryk Brixi
Garyk Brixi
MDMatthew G. Durrant
Matthew G. Durrant
JKJerome Ku
Jerome Ku
MNMohsen Naghipourfar
Mohsen Naghipourfar
MPMichael Poli
Michael Poli
BHBrian L. Hie
Brian L. Hie
All of life encodes information with DNA. Although tools for genome sequencing, synthesis and editing have transformed biological research, we still lack sufficient understanding of the immense complexity encoded by genomes to predict the effects of many classes of genomic changes or to intelligently compose new biological systems. Artificial intelligence models that learn information from genomic sequences across diverse organisms have increasingly advanced prediction and design capabilities1,2. Here we introduce Evo 2, a biological foundation model trained on 9 trillion DNA base pairs from a highly curated genomic atlas spanning all domains of life to have a 1 million token context window with single-nucleotide resolution. Evo 2 learns to accurately predict the functional impacts of genetic variation—from noncoding pathogenic mutations to clinically significant BRCA1 variants—without task-specific fine-tuning. Mechanistic interpretability analyses reveal that Evo 2 learns representations associated with biological features, including exon–intron boundaries, transcription factor binding sites, protein structural elements and prophage genomic regions. The generative abilities of Evo 2 produce mitochondrial, prokaryotic and eukaryotic sequences at genome scale with greater naturalness and coherence than previous methods. Evo 2 also generates experimentally validated chromatin accessibility patterns when guided by predictive models3,4 and inference-time search. We have made Evo 2 fully open, including model parameters, training code5, inference code and the OpenGenome2 dataset, to accelerate the exploration and design of biological complexity.

7 September 2024 in bioRxiv
Verified
Preprint
Open access
PDF file
YPYash Patel
Yash Patel
CZChenghao Zhu
Chenghao Zhu
TYTakafumi N. Yamaguchi
Takafumi N. Yamaguchi
NWNicholas K. Wang
Nicholas K. Wang
NWNicholas Wiltsie
Nicholas Wiltsie
PBPaul C. Boutros
Paul C. Boutros
Summary DNA sequencing is becoming more affordable and faster through advances in high-throughput technologies. This rise in data availability has contributed to the development of novel algorithms to elucidate previously obscure features and led to an increased reliance on complex workflows to integrate such tools into analyses pipelines. To facilitate the analysis of DNA sequencing data, we created metapipeline-DNA, a highly configurable and extensible pipeline. It encompasses a broad range of processing including raw sequencing read alignment and recalibration, variant calling, quality control and subclonal reconstruction. Metapipeline-DNA also contains configuration options to select and tune analyses while being robust to failures. This standardizes and simplifies the ability to analyze large DNA sequencing in both clinical and research settings. Availability Metapipeline-DNA is an open-source Nextflow pipeline under the GPLv2 license and is freely available at https://github.com/uclahs-cds/metapipeline-DNA.

11 April 2026 in bioRxiv
Verified
Preprint
Open access
PDF file
MPMichael T Pearce
Michael T Pearce
TDThomas Dooms
Thomas Dooms
RYRyo Yamamoto
Ryo Yamamoto
SAShant Ayanian
Shant Ayanian
ARAlexander Ryu
Alexander Ryu
NWNicholas K Wang
Nicholas K Wang
Scientific foundation models learn high-dimensional representations from diverse data modalities, yet what they encode and how to extract that knowledge remain open questions. Here we show that probing the internal representations of Evo 2, a 7-billion-parameter genomic foundation model, enables accurate and interpretable genetic variant effect prediction. We introduce a covariance-based probe that captures second-order structure from Evo 2 sequence embeddings to predict variant pathogenicity across variant types and functional consequences, matching or exceeding specialized predictors within their domains. To ground these predictions in known biological mechanisms, we train a complementary panel of probes on existing annotations to detect which genomic properties are disrupted by a variant. This categorized evidence is then integrated with each variant's genomic context through a language model to generate variant-specific mechanistic hypotheses. Our pathogenicity predictions correlate with experimental measures of variant function, clinical penetrance, and biobank disease associations while the mechanistic hypotheses are consistent with expert reviews, known mechanism classes, and downstream molecular readouts. We release pathogenicity scores, disruption profiles, and contextualized interpretations for 4.2 million variants from the ClinVar database as an open resource through the Evo Variant Effect Explorer (EVEE). More broadly, this structured probing approach offers a general framework for interrogating foundation models across scientific disciplines and grounding their outputs in existing domain concepts.

3 November 2024 in bioRxiv
Verified
Preprint
Open access
PDF file
NWNicholas K. Wang
Nicholas K. Wang
NWNicholas Wiltsie
Nicholas Wiltsie
HWHelena K. Winata
Helena K. Winata
SFSorel Fitz-Gibbon
Sorel Fitz-Gibbon
AGAlfredo E. Gonzalez
Alfredo E. Gonzalez
PBPaul C. Boutros
Paul C. Boutros
Reference genomes are foundational to modern genomics. Our growing understanding of genome structure leads to continual improvements in reference genomes and new genome “builds” with incompatible coordinate systems. We quantified the impact of genome build on germline and somatic variant calling by analyzing tumour-normal whole-genome pairs against the two most widely used human genome builds. The average individual had a build-discordance of 3.8% for germline SNPs, 8.6% for germline SVs, 25.9% for somatic SNVs and 49.6% for somatic SVs. Build-discordant variants are not simply false-positives: 47% were verified by targeted resequencing. Build-discordant variants were associated with specific genomic and technical features in variant- and algorithm-specific patterns. We leveraged these patterns to create StableLift, an algorithm that predicts cross-build stability with AUROCs of 0.934 ± 0.029. These results call for significant caution in cross-build analyses and for use of StableLift as a computationally efficient solution to mitigate inter-build artifacts.