Pith. sign in

REVIEW 3 major objections 6 minor 25 references

Minimal readout probes on Evo 2 embeddings detect antibiotic-resistance genes with high accuracy.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-02 02:50 UTC pith:VYQYWFWT

load-bearing objection A competent, honest probing study showing Evo 2 features linearly decode AMR/virulence at CDS level; the central result is plausible, but the pretraining-overlap question and withheld code/data keep it conditional. the 3 major comments →

arxiv 2607.14070 v1 pith:VYQYWFWT submitted 2026-07-15 q-bio.GN cs.LG

Screening of Biosecurity Features in Metagenomic Data with Evo 2 Probes

Jeremy Guntoro , Alexander Dack , Dylan Danno , Michaela Jan\v{c}ovi\v{c}ov\'a , Kri\v{z}an Jurinovi\'c , Vanessa Smilansky This is my paper
classification q-bio.GN cs.LG
keywords genomic foundation modelsEvo 2 embeddingslinear probingattention probingantimicrobial resistance detectionmetagenomic biosurveillancebacterial virulenceshort-read screening
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper tries to establish that biosecurity-relevant information is already present, in linearly readable form, inside the frozen representations of a large DNA language model. The authors train only a single lightweight layer on top of Evo 2's internal per-nucleotide vectors and show it distinguishes antibiotic-resistance regions from matched non-resistant regions across held-out metagenomes, reaching ROC-AUC 0.888 with mean-pooling and 0.977 with attention pooling. The same probe separates AMR drug-class subcategories, detects bacterial virulence more weakly (0.833), and keeps its ranking performance on simulated short sequencing reads (0.898) without retraining. If this holds, embedding-based probes could serve as a fast, cheap first-pass screen that runs before genome assembly. The paper is careful about one limit: simulated reads are an optimistic proxy for real samples, and prompt labels on AI-generated sequences do not prove the sequences function as resistance genes.

Core claim

Frozen Evo 2 layer-26 activations contain linearly decodable biosecurity signal. A mean-pooled linear probe over per-token activations reaches region-level ROC-AUC 0.888 for antimicrobial resistance (AMR) on held-out metagenome-assembled genomes; a single-head attention probe reaches 0.977 and localizes its contribution to the annotated resistance gene. Per-drug-class probes separate five AMR classes one-vs-rest with in-distribution AUC above 0.98 and reject unrelated functional genes almost as well (the beta-lactam probe drops to 0.783 against virulence factors). The unchanged linear probe scores simulated short reads at ROC-AUC 0.898, comparable to its full-region result. Bacterial virulen

What carries the argument

The central object is the layer-26 residual-stream activation of Evo 2, a 4096-dimensional vector per DNA token, taken frozen and used as input to a single-layer probe. The linear probe mean-pools per-token logits; the attention probe learns a query that weights informative tokens. The load-bearing work is that a single uniformly-pooled linear map over these activations already separates AMR from length- and GC-matched negatives drawn from the same genome, and that the signal survives the shift to short noisy reads.

Load-bearing premise

The read-level screening claim assumes the simulator's MiSeq error and coverage model captures how real metagenomic reads diverge from assembled contigs; the authors themselves call simulated reads an optimistic proxy.

What would settle it

Take the identical frozen-embedding probe and score genuine shotgun metagenomic reads from samples whose AMR content is confirmed by independent methods (culture, PCR, or assembly-based annotation). If the read-level ROC-AUC on authentic data falls to chance, the before-assembly screening use case is not supported.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

Share X Bluesky LinkedIn Reddit HN

If this is right

  • AMR ranking transfers to simulated short reads without retraining, suggesting screening could be run before assembly in low-coverage or strain-heterogeneous samples.
  • Because the base model stays frozen, probes for many resistance classes and other hazards can be trained and applied in parallel at small marginal cost.
  • Attention pooling improves region-level AMR discrimination and pinpoints the resistance gene more sharply than uniform mean-pooling.
  • Drug-class-specific AMR signals are separable from each other and from unrelated functional genes, evidence the probe is not just detecting generic coding status.
  • Weak recoverability of AMR prompt labels in generated sequences means AI-generated sequence prompts should not be treated as functional labels without wet-lab verification.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If real metagenomic reads behave like the simulation, this could become the triage layer of a biosurveillance pipeline: flag reads cheaply, assemble only flagged samples, and reserve alignment-based confirmation for positives.
  • The same probing recipe likely extends to other biosecurity-relevant elements, such as toxins, virulence cassettes, or engineered resistance markers in synthetic-DNA orders.
  • Since the attention probe beat the linear probe on AMR, applying attention pooling to virulence, drug-class, and read-level tasks could close the gap there too; a short-read-trained sparse autoencoder might restore some of the lost interpretability.
  • The SynGenome result is a caution for generative genomics: prompt-derived labels can leak into generated sequences only weakly, so measured function, not prompt association, is the right target for risk assessment.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

3 major / 6 minor

Summary. The paper trains minimal linear and single-head attention probes on frozen Evo 2 layer-26 activations to detect biosecurity-relevant features in metagenomic data: AMR genes, AMR drug-class subcategories, and bacterial virulence. The evaluation uses held-out MAG-level (AMR) and species-level (VFDB) splits with matched negatives, and the AMR probe is further applied to InSilicoSeq-simulated short reads. The central quantitative results are region-level ROC-AUC 0.888 for the linear AMR probe, 0.977 for the attention probe, 0.898 for the linear probe on simulated reads, and 0.833 for bacterial virulence. The paper also reports a weak SynGenome prompt-label result (AUC 0.599) and an exploratory sparse-autoencoder analysis. The authors conclude that embedding-based probes are a promising cheap first-pass screening layer for metagenomic biosurveillance.

Significance. If the results hold, the paper provides a useful demonstration that simple probes on a genomic foundation model can recover functional annotations such as AMR and virulence with strong discrimination, and that the read-level transfer works in simulation. The design has genuine strengths: splits by MAG/species to reduce leakage, matched negative controls for length and GC content, transfer to per-drug-class and functional-category negative pools, and honest reporting of weak or exploratory results (SynGenome AUC 0.599, SAE reconstruction gap, small transfer-negative sets). However, the central generalization claim rests on the assumption that the test MAGs/species are not memorized by Evo 2 during pretraining, and this is not checked. The lack of confidence intervals on all reported AUC/F1 values, and the confounded attention-vs-linear comparison, further limit the strength of the quantitative claims. With additional analyses the result could be made convincing for a journal; in its current form it is a promising conditional result.

major comments (3)
  1. [Section 3.2 and Section 5, Table 2] The held-out MAG/species splits prevent leakage between probe training and evaluation, but they do not address possible overlap between test sequences and Evo 2's pretraining corpus. MGnify and VFDB reference genomes are exactly the kind of data likely to be in a tree-of-life training set, and if the test MAGs are highly similar to pretraining sequences, the probes may read out sequence-specific memory rather than generalizable functional representations. This directly affects the central claim that Evo 2 'encodes these functional properties' in a way useful for screening novel or engineered threats. Please add a contamination analysis: quantify exact/near-exact or k-mer-level overlap between the test regions and Evo 2 pretraining data, or demonstrate transfer to genuinely novel sequences (e.g., new assemblies or validated synthetic constructs). The weak SynGenome transfer (AUC 0.599) ma
  2. [Tables 2 and 3] No confidence intervals or uncertainty estimates are reported for any AUC or F1. This matters for several load-bearing comparisons: the attention-probe advantage over the linear probe (0.977 vs 0.888), the read-level vs region-level comparison (0.898 vs 0.888), and the per-drug-class transfer results, where the transfer-negative pools are tiny (STRESS n=47, VIRULENCE n=31) and the beta-lactam vs VIRULENCE AUC is 0.783. Without bootstrap or equivalent intervals, it is impossible to tell which differences are meaningful. Please add CIs or a similar uncertainty quantification, at least for the headline numbers and for the beta-lactam/virulence transfer row.
  3. [Section 3.3, Section 4.1.1, and Conclusion] The attention probe is trained with a region-level objective while the linear probe is trained with per-token supervision, so the comparison conflates architecture with training objective. The paper acknowledges this in the conclusion, but the abstract and Results state the 0.977 vs 0.888 difference without the caveat. To support the claim that attention pooling itself improves detection, either train an attention probe with token-level supervision, or train a linear probe with region-level supervision, and compare like with like. At minimum, move the caveat into the Results section where the comparison is first made.
minor comments (6)
  1. [Abstract and Section 4.1.1] The phrase 'enabling evaluation before assembly' overstates what simulated reads demonstrate; the Discussion correctly calls simulated reads 'an optimistic proxy.' Please qualify the abstract claim (e.g., 'potentially enabling') and explicitly state that real-read validation is missing.
  2. [Figure 1 caption] The organism name 'Onthenecus intestinigallinarum' appears misspelled; please verify and correct.
  3. [Section 3.3 / Table A3] The pos_weight=20.0 hyperparameter for the AMR binary probe is a free parameter, but its sensitivity is not analyzed. A short robustness check (e.g., pos_weight in {5,10,20,50}) would strengthen the claim that performance is not highly threshold- or weighting-dependent.
  4. [Section 4.1.4 and Figure 2] The SAE feature-selection analysis is explicitly exploratory and performed on the same data, which is good to state, but the figure lacks error bars or a statistical test. Adding standard errors across regions (or a permutation test) would make the descriptive claim more informative.
  5. [Code and Data Availability] The lack of publicly released code/data limits reproducibility. The 'reasonable request' policy is understandable for dual-use reasons, but the paper should at least provide a more detailed pseudocode or an appendix containing the exact split definitions, matching code inputs, and hyperparameter search ranges so that the results can be reconstructed independently.
  6. [Section 4.1.1] The text says the read-level F1 threshold is selected on one randomly sampled clean 301 bp segment per validation CDS. Please clarify how many validation CDS were used and whether the random sampling was seeded/repeated; otherwise, the threshold calibration has an unnecessary source of variance.

Circularity Check

0 steps flagged

No significant circularity: held-out probe evaluations are self-contained; the only same-dataset SAE exploration is explicitly labeled descriptive.

full rationale

The central derivations are not circular. AMR, attention, and virulence probes are trained on frozen Evo 2 layer-26 activations with external labels and evaluated on held-out MAGs/species that are disjoint by split unit: the paper states 'no MAG or species is shared between train, val and test, so evaluation is always on unseen MAGs (for AMR) or species (for BV)'. The reported AUCs are measured on test regions the probes never saw. The simulated-read result applies the unchanged trained probe to InSilicoSeq reads with a threshold calibrated on validation fragments, making it a genuine transfer measurement rather than a refit. Per-drug-class probes are trained one-vs-rest and tested against separate NEG/STRESS/VIRULENCE pools. The SynGenome experiment is honestly labeled as weak and as not establishing function: 'these prompt-derived labels do not establish the function of the generated response sequences.' The only same-dataset component is the SAE feature inspection, which the paper explicitly disclaims: 'Because feature selection and visualisation were performed on the same dataset, the resulting dimensions should be interpreted as descriptive candidate associations rather than independently validated features.' This is a labeled exploration, not a hidden prediction. The layer-26 choice is motivated by prior external work (Brixi et al., Gorton et al.) as a hypothesis, and none of the cited works are by the present authors; there is no load-bearing self-citation chain. The concern that Evo 2 may have seen test genomes during pretraining is a real external-validity/memorization risk, but it is not a definitional reduction of the prediction to its input, so it does not constitute circularity under the rubric.

Axiom & Free-Parameter Ledger

1 free parameters · 5 axioms · 0 invented entities

The paper is an empirical ML study. Trained probe weights are the fitted objects and are evaluated on held-out MAGs/species, so they are not circular theory parameters. The only hand-chosen training constant listed is the positive-token upweight; F1-maximising thresholds are validation-selected but do not affect the AUC-based central claims. The claims rest on five domain assumptions covering label fidelity, matched-negative sufficiency, simulation realism, and pretraining overlap. No invented entities are introduced.

free parameters (1)
  • pos_weight (AMR binary probe positive-token upweight) = 20.0
    Hand-chosen training weight for rare positive tokens in the AMR binary probe (Appendix A, Training configuration). It affects the trained model but not the AUC evaluation protocol; listed for completeness.
axioms (5)
  • domain assumption Evo 2-7B-262k layer-26 activations are general-purpose features that transfer to metagenomic MAG and short-read inputs
    The method freezes Evo 2 and reads layer-26 activations (Section 3.1); if the pretrained representation does not transfer to metagenomic/read-length inputs, the probe results do not generalize.
  • domain assumption MGnify AMR annotations and VFDB labels correctly identify resistant and virulent CDS
    Labels come from curated databases without wet-lab verification (Section 3.2, Table 1); label noise or misannotation propagates into the reported AUCs.
  • domain assumption InSilicoSeq MiSeq error model simulates realistic metagenomic short reads
    Read-level transfer (Section 3.4) assumes the simulated reads capture real sequencing error, GC bias, and fragmentation; the Discussion calls simulated reads 'an optimistic proxy'.
  • domain assumption Matched negatives remove generic functional-gene and taxonomic confounds
    The AMR-specificity claim rests on same-MAG/same-species matching with length/GC constraints and exclusion of other functional categories (Section 3.3); residual leakage would inflate transfer AUCs.
  • domain assumption Evo 2's pretraining corpus does not overlap the evaluation genomes in a way that inflates probe AUCs
    'Held-out' refers to probe training; Evo 2 was pretrained on large genomic databases, and the paper never checks whether MGnify test MAGs or close homologs were seen during pretraining.

pith-pipeline@v1.3.0-alltime-deepseek · 14051 in / 15264 out tokens · 138387 ms · 2026-08-02T02:50:22.255939+00:00 · methodology

0 comments
Cite this review

Pith. "Pith review of Screening of Biosecurity Features in Metagenomic Data with Evo 2 Probes." pith.science (2026). https://pith.science/paper/VYQYWFWT

@misc{pith2026260714070,
  author       = {Pith},
  title        = {Pith review of: Screening of Biosecurity Features in Metagenomic Data with Evo 2 Probes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/VYQYWFWT}},
  note         = {Machine review of arXiv:2607.14070}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Genomic foundation models such as Evo 2 learn rich sequence representations, but their value for biosecurity screening is largely unexplored. We ask how much biosecurity-relevant signal is linearly accessible in these representations by training minimal linear and attention probes on frozen Evo 2 layer-26 activations, without fine-tuning the underlying model. Across held-out metagenomic test sets, the probes detect antimicrobial resistance (AMR) with strong discrimination: a linear probe reaches a region-level ROC-AUC of 0.888 (mean-pool), rising to 0.977 with a single-head attention probe. The probes resolve finer-grained AMR drug-class subcategories and separate them from unrelated functional genes, providing additional evidence that the learned signal is not explained solely by generic functional-gene status. Bacterial virulence is also decodable, though more weakly (region-level ROC-AUC 0.833). The AMR probe retains comparable ranking performance on simulated short reads without retraining, enabling evaluation before assembly in settings where assembly is computationally costly or unreliable. It achieves a read-level ROC-AUC of 0.898 (mean-pool), comparable to the mean-pooled full-region result. Within SynGenome, AMR-associated prompt labels are only weakly recoverable from Evo 1.5-generated sequences; these prompt-derived labels do not establish the function of the generated response sequences. A complementary sparse-autoencoder analysis recovers interpretable resistance-associated features but proves less consistent than the supervised probes. Together, these results position lightweight embedding-based probes as a fast, inexpensive first-pass detection layer for metagenomic biosurveillance and map both strengths and current limits of the approach. This work was conducted as part of the AIxBio Hackathon 2026 hosted by BlueDot Impact, Apart Research, and Cambridge Biosecurity Hub.

Figures

Figures reproduced from arXiv: 2607.14070 by Alexander Dack, Dylan Danno, Jeremy Guntoro, Kri\v{z}an Jurinovi\'c, Michaela Jan\v{c}ovi\v{c}ov\'a, Vanessa Smilansky.

Figure 1
Figure 1. Figure 1: Mean-pooled distributions of per-region linear probe scores across datasets. The linear probe produces a score at each token (i.e., nucleotide position), which is then mean-pooled over annotated genomic regions (including flanking regions where present and coding sequences) to obtain a single score per region; these region-level scores are aggregated into distributions over positive and negative regions. (… view at source ↗
Figure 2
Figure 2. Figure 2: Exploratory comparison of sparse autoencoder features associated with macrolide antibiotic resistance in the AMR-binary dataset. We use a pre-trained SAE from Goodfire (Brixi et al., 2026; Gorton et al., 2025). Two salient features (SAE latent di￾mension) are plotted on the x-axis, with two bars per feature cor￾responding to different subsets of genomic regions. The SAE is applied to token-level embeddings… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

25 extracted references · 3 linked inside Pith

  1. [1]

    2016 , month = oct, eprint =

    Alain, Guillaume and Bengio, Yoshua , title =. 2016 , month = oct, eprint =

    2016

  2. [2]

    and Raphenya, Amogelang R

    Alcock, Brian P. and Raphenya, Amogelang R. and Lau, Tammy T. Y. and Tsang, Kara K. and Bouchard, M. Nucleic Acids Research , volume =. 2020 , month = jan, doi =

    2020

  3. [3]

    and Vikesland, Peter and Zhang, Liqing , title =

    Arango-Argoty, Gustavo and Garner, Emily and Pruden, Amy and Heath, Lenwood S. and Vikesland, Peter and Zhang, Liqing , title =. Microbiome , volume =. 2018 , month = dec, doi =

    2018

  4. [4]

    and Ku, Jerome and Naghipourfar, Mohsen and Poli, Michael and Sun, Gwanggyu and Brockman, Greg and Chang, Daniel and Fanton, Alison and Gonzalez, Gabriel A

    Brixi, Garyk and Durrant, Matthew G. and Ku, Jerome and Naghipourfar, Mohsen and Poli, Michael and Sun, Gwanggyu and Brockman, Greg and Chang, Daniel and Fanton, Alison and Gonzalez, Gabriel A. and others , title =. Nature , volume =. 2026 , month = mar, doi =

    2026

  5. [5]

    and Sirelkhatim, Hassan and Richard, Guillaume and Skwark, Marcin and Beguir, Karim and Lopez, Marie and Pierrot, Thomas , title =

    Dalla-Torre, Hugo and Gonzalez, Liam and Mendoza-Revilla, Javier and Lopez Carranza, Nicolas and Grzywaczewski, Adam Henryk and Oteri, Francesco and Dallago, Christian and Trop, Evan and De Almeida, Bernardo P. and Sirelkhatim, Hassan and Richard, Guillaume and Skwark, Marcin and Beguir, Karim and Lopez, Marie and Pierrot, Thomas , title =. Nature Methods...

    2025

  6. [6]

    and Haendiges, Julie and Haft, Daniel H

    Feldgarden, Michael and Brover, Vyacheslav and Gonzalez-Escalona, Narjol and Frye, Jonathan G. and Haendiges, Julie and Haft, Daniel H. and Hoffmann, Maria and Pettengill, James B. and Prasad, Arjun B. and Tillman, Glenn E. and Tyson, Gregory H. and Klimke, William , title =. Scientific Reports , volume =. 2021 , month = jun, doi =

    2021

  7. [7]

    2025 , month = apr, day =

    Fields, Evan and Bhasin, Harmon and Teo, Ryan and McLaren, Mike and Kaufman, Jeff , title =. 2025 , month = apr, day =

    2025

  8. [8]

    , title =

    Florensa, Alfred Ferrer and Kaas, Rolf Sommer and Clausen, Philip Thomas Lanken Conradsen and Aytan-Aktug, Derya and Aarestrup, Frank M. , title =. Microbial Genomics , volume =. 2022 , month = jan, doi =

    2022

  9. [9]

    2025 , month = feb, howpublished =

    Gorton, Liv and Wang, Nicholas and Nguyen, Nam and Deng, Myra and Ho, Eric and Balsam, Daniel and McGrath, Thomas , title =. 2025 , month = feb, howpublished =

    2025

  10. [10]

    Simulating

    Gourl. Simulating. Bioinformatics , volume =. 2019 , month = feb, doi =

    2019

  11. [11]

    Registry of Standard Biological Parts , year =

  12. [12]

    and Wittmann, Bruce J

    Ikonomova, Svetlana P. and Wittmann, Bruce J. and Piorino, Fernanda and Ross, David J. and Schaffter, Samuel W. and Vasilyeva, Olga and Horvitz, Eric and Diggans, James and Strychalski, Elizabeth A. and Lin-Gibson, Sheng and Taghon, Geoffrey J. , title =. bioRxiv , year =

  13. [13]

    , title =

    Ji, Yanrong and Zhou, Zhihan and Liu, Han and Davuluri, Ramana V. , title =. Bioinformatics , volume =. 2021 , month = aug, doi =

    2021

  14. [14]

    Proceedings of the 42nd International Conference on Machine Learning , year =

    Kantamneni, Subhash and Engels, Joshua and Rajamanoharan, Senthooran and Tegmark, Max and Nanda, Neel , title =. Proceedings of the 42nd International Conference on Machine Learning , year =. 2502.16681 , archiveprefix =

    Pith/arXiv arXiv

  15. [15]

    and Driscoll, Claudia L

    King, Samuel H. and Driscoll, Claudia L. and Li, David B. and Merchant, Aditi T. and Wilkinson, Max E. and Hie, Brian L. and others , title =. bioRxiv , year =

  16. [16]

    and Ba, Jimmy , title =

    Kingma, Diederik P. and Ba, Jimmy , title =. 3rd International Conference on Learning Representations , year =. 1412.6980 , archiveprefix =

    Pith/arXiv arXiv

  17. [17]

    Nucleic Acids Research , volume =

    Liu, Bo and Zheng, Dandan and Zhou, Siyu and Chen, Lihong and Yang, Jian , title =. Nucleic Acids Research , volume =. 2022 , doi =

    2022

  18. [18]

    2025 , month = jan, eprint =

    Liu, Ollie and Jaghouar, Sami and Hagemann, Johannes and Wang, Shangshang and Wiemels, Jason and Kaufman, Jeff and Neiswanger, Willie , title =. 2025 , month = jan, eprint =

    2025

  19. [19]

    Advances in Neural Information Processing Systems , year =

    McKenzie, Alex and Pawar, Urja and Blandfort, Phil and Bankes, William and Krueger, David and Lubana, Ekdeep Singh and Krasheninnikov, Dmitrii , title =. Advances in Neural Information Processing Systems , year =. 2506.10805 , archiveprefix =

    arXiv

  20. [20]

    and King, Samuel H

    Merchant, Aditi T. and King, Samuel H. and Nguyen, Eric and Hie, Brian L. , title =. Nature , volume =. 2026 , doi =

    2026

  21. [21]

    Nguyen, Eric and Poli, Michael and Faizi, Marjan and Thomas, Armin and Birch-Sykes, Callum and Wornow, Michael and Patel, Aman and Rabideau, Clayton and Massaroli, Stefano and Bengio, Yoshua and Ermon, Stefano and Baccus, Stephen A. and R. Advances in Neural Information Processing Systems , year =. 2306.15794 , archiveprefix =

    Pith/arXiv arXiv

  22. [22]

    and Kang, Brian and Katrekar, Dhruva and Li, David B

    Nguyen, Eric and Poli, Michael and Durrant, Matthew G. and Kang, Brian and Katrekar, Dhruva and Li, David B. and Bartie, Liam J. and Thomas, Armin W. and King, Samuel H. and Brixi, Garyk and Sullivan, Jeremy and Ng, Madelena Y. and Lewis, Ashley and Lou, Aaron and Ermon, Stefano and Baccus, Stephen A. and Hernandez-Boussard, Tina and R. Sequence Modeling ...

    2024

  23. [23]

    Pearce, Michael T. and Dooms, Thomas and Yamamoto, Ryo and Meehl, Joshua and Molnar, Carl and Bissell, Mark and Hazra, Dron and Fang, Ching and Nguyen, Nam and Anderson, Michael and Osborne, Collin and Duffy, Patrick and Toomey, Bridget and Klee, Eric and Myasoedova, Elena and Ryu, Alexander J. and Ayanian, Shant and Korfiatis, Panos and Redlon, Matt and ...

  24. [24]

    and Burdett, Tony and Burgin, Josephine and Caballero-P

    Richardson, Lorna and Allen, Ben and Baldi, Germana and Beracochea, Martin and Bileschi, Maxwell L. and Burdett, Tony and Burgin, Josephine and Caballero-P. Nucleic Acids Research , volume =. 2023 , doi =

    2023

  25. [25]

    2026 , month = mar, eprint =

    Tai, Huilin , title =. 2026 , month = mar, eprint =

    2026