Skip to content
OctymPowered By Proofig AI
Research integrity & quality platform · from the team behindProofig AI

Every integrity and quality check, in one review. Configured to your needs.

One platform for researchers, institutions, and publishers, from grant application to final publication. It catches problems at the moment they are easiest to address, and checks every manuscript against your own requirements: your journal's, your funder's, your policies.

  • From grant application to publication
  • Every finding traced to its source
  • Configured by your own Octym team
J. Genome Maint. Res. · Vol 12 · Issue 4 · 2026doi:10.5555/jgmr.2026.04.138Received 12 Jan · Revised 9 Mar · Accepted 28 Mar 2026 · Open AccessRAD51–BRCA2 coupling stabilises replicationforks under oncogene-induced stressA. Thornton, M. Delgado, P. Ramaswamy, J. O'Brien, L. Hartmann, H. Tanaka¹Example University · ²Example Institute of Cell Biology · ³Example Genome Center · *[email protected]AbstractReplication stress is a hallmark of oncogene-driven tumorigenesis, yet how RAD51 and BRCA2 stabilise stalledforks in primary human cells remains incompletely resolved. We apply longitudinal single-cell imaging (n = 4,812cells, 23 donors) to quantify fork restart kinetics under hydroxyurea challenge, and show RAD51 residence timescales with BRCA2 occupancy (R² = 0.84, p < 0.001), a coupling disrupted in BRCA2-L2510P carriers.Keywords: replication stress · homologous recombination · RAD51 · BRCA2 · single-cell imaging1. IntroductionThe integrity of replication forks under oncogenic stressis a central fulcrum of genome stability [1,2]. RAD51nucleofilament assembly is rate-limiting for protection[3], and its loading by BRCA2 is a canonical tumour-suppressive axis [5,6]. Prior bulk assays gave inconsistentRAD51:BRCA2 ratios for fork stability, motivating thesingle-cell, single-molecule approach developed here.2. ResultsWe established a longitudinal imaging pipeline in CD34⁺progenitors, tracking endogenous RAD51-mEGFP and BRCA2-mCherry across 72 h of hydroxyurea exposure (Fig. 1).Single-molecule photobleaching confirmed 1:1 labelling;fork-restart frequency correlated with mean RAD51residence time (τ_res = 18.3 ± 2.1 s, Fig. 2).250150100755037RAD51BRCA2γH2AXα-tubulinCtrlWTWT+HUKOL2510PFig. 1. Immunoblot of iPOND fork complexes; co-enrichmentof RAD51/BRCA2 is lost in L2510P carriers (lane 5).2.1 Single-molecule trackingImaging used a Nikon Ti2 with a 60× 1.49 NA TIRFobjective at 50 ms exposure, 10 Hz, over 600 s per field.Trajectories were linked in TrackMate v7.11 and filteredfor SNR > 4. Restart latencies were bimodal: a fastpopulation (τ₁ = 41 ± 6 s) comprised 68% of events, theslow class enriched in monoallelic BRCA2 carriers.2.2 Fork-complex compositioniPOND immunoblots showed stoichiometric co-enrichment ofRAD51 and BRCA2 in wild-type cells (ratio 1.07 ± 0.14),falling to 0.31 ± 0.08 in L2510P (p < 0.001). This losstracked with accelerated fork degradation and elevatedγH2AX, indicating compromised fork protection. DNA combingconfirmed reduced fork speed and symmetry in carriers.3. MethodsCD34⁺ cells from three donors (IRB #2026-0418)were cultured in StemSpan SFEM II with SCF, FLT3L and TPO.Hydroxyurea was applied at 2 mM for the durations shown.Foci were segmented with Cellpose and quantified pernucleus; comparisons used two-way ANOVA with Bonferronicorrection. The fork-restart rate constant was derived as:krest=[BRCA2] · τK_d + [BRCA2](1)20406080102540557085100Time after HU (min)RAD51 foci / nucleusControl (n=23)L2510P (n=18)***Fig. 2. Focus-resolution kinetics; L2510P shows acceleratedloss (p < 0.001, two-way ANOVA, error bars = SEM).Confocal imaging of nuclear foci revealed pronouncedspatial co-localisation of RAD51 and γH2AX at stalledforks (Fig. 3; Pearson r = 0.78, n = 412 fields), largelyabsent in cells lacking functional BRCA2.5µmDAPI5µmRAD515µmγH2AX5µmMergeFig. 3. Confocal RAD51/γH2AX foci; scale bar 5 µm.Table 1. Summary statistics (mean ± SD).GenotypenfociτpWT239.141–Het217.473.02L2510P185.3112<.001Data: data.example.org/record/10918842References[1] Okafor N, Lindqvist E.Sources of replication stress in dividing cells. Replication Rev (2019).[2] Brandt H, Ishikawa T, et al.Clamp loading at stalled replication forks. Fork Biol Rep (2012).[3] Moreau C, Adeyemi K, et al.Fork protection in repair-deficient cells. J Genome Maint Res (2018).[4] Varga L, Sato Y, et al.Nascent strand protection at stalled forks. Chromatin Stab Lett (2015).[5] Halvorsen P, Mehta R, et al.Recombinase loading at stalled forks. Recomb Biol Commun (2010).[6] Castell D, Novak J, et al.Fork reversal in repair-defective cells. Replication Rev (2021).AI-generated textAI-generated textImage manipulationImage manipulationImage duplicationImage duplicationAI-generated imagesAI-generated imagesGraph & chart duplicationGraph & chart duplicationStatistical consistencyStatistical consistencyFigure–legend matchFigure–legend matchIncomplete referencesIncomplete referencesImage plagiarismImage plagiarismRetracted referencesRetracted referencesHallucinated referencesHallucinated referencesData availabilityData availabilityAuthorship & affiliationsAuthorship & affiliationsPaper-mill signalsPaper-mill signalsReference analyticsReference analyticsResearch ethicsResearch ethicsMethods & rigorMethods & rigorNovelty & contributionNovelty & contributionPeer-review analysisPeer-review analysisReproducibilityReproducibilityReplicability signalsReplicability signalsIncomplete referencesMethods & rigorStatistical consistencyClaim–citation consistencyImage manipulationRetracted references
The shift

Today, every check lives in a different tool.

Images in one place, references in another, statistics in a third: each a signal with no context, and the problems that cause rejections and retractions surface too late. Octym replaces all of it with one review.

surfaced too lateOCTYM · ONE REVIEWtracedtracedtracedtracedtracedtraced

Six tools, six signals, collapsed into one review

Tailor-made

Your requirements. Your rules. Configured for you.

At Octym, you don't just get the platform. You get a team. People who learn your rules, your requirements, and what you're most worried about; who set up your review to match; and who change it whenever you need something new.

A journal's screen, a funder's checklist, an investigation's depth. Built for you, by people you can actually talk to.

OCTYM · YOUR REVIEW

You ask · your Octym team shapes it

What we check

Every check that decides whether a manuscript holds up.

Grouped into three categories: what's inside the manuscript, how it stands against the published record and the requirements of wherever it's headed, and whether the science is sound. Each weighted equally, every finding traced to its source.

One review · many lenses

Internal manuscript integrity

AI-generated text signalsImage manipulationImage duplicationAI-generated imagesGraph & chart duplicationStatistical & analytical consistencyFigure–legend consistencyClaim & citation consistencyInternal consistencyPrior publication & reuse disclosure

External checks & requirements

Image & graph plagiarismRetracted reference detectionHallucinated / non-existent referencesIncomplete & broken referencesReference analyticsData & code availabilityPaper-mill signalsAuthorship & affiliation consistencyFunder requirementsSubmission requirements & completeness

Research quality

Methods & rigorReproducibilityReplicability signalsResearch ethicsNovelty & contributionPeer-review-grade analysis

Under the surface

Scrutinized from every side. Until it holds.

Weighed against the rest of the paper, the data behind it, and the published record.

The difference

What only Octym brings.

Talk to your reviewOnly on Octym

A review you can talk to.

Ask the agent anything: why a figure was flagged, where the reproducibility gaps are, whether the ethics statement holds. All answered from the manuscript, every claim linked to its source.

Fig. 2p = 0.012[14] Marsh et al., 2021…
Octym agentonline

Click a finding to ask about it…

Ask anything about this manuscript…
Image & figure integrityPowered by Proofig AI

Image integrity, at the depth the science demands.

Powered by Proofig AI (adopted by the Science family of journals and trusted by the world's top research institutions and publishers), with forensics that compare against the published literature and also detect duplication and plagiarism in graphs, charts, diagrams, and plots.

Reference & citation intelligencePowered by RefGuard

Every reference, checked against the record.

Retracted, hallucinated, and mismatched references, and citations that don't support their claim: each shown in context, with the analytics behind the bibliography.

ReferencesAll 58Retracted 1Hallucinated 1Unsupported 3
  • 14

    Zhang et al., 2019

    Retraction notice, 2022

    Retracted
  • 22

    Okafor & Lee, 2021 (DOI not found)

    Hallucinated
  • 31

    Garcia et al., 2018

    Cited for X; reports Y

    Doesn't support claim

Reference age

Open access

0%

Top venue

0%

From the team behind Proofig

The most trusted name in scientific image integrity, now reviewing the whole manuscript.

Octym is built by the team behind Proofig, and Proofig itself powers the image-integrity checks inside every Octym review.

Adopted by Science

The Science family of journals uses Proofig to screen images across all six of its titles.

Proofig

Trusted across research

Relied on by the world's top researchers, publishers, and research institutions.

Proofig

The largest image-plagiarism database

Every figure checked by Proofig for manipulation, duplication, plagiarism, and AI generation.

Proofig
Watch

Octym, in sixty seconds.

Every integrity and quality check, in one review. See how it works.

See it on any manuscript.

Contact us, or log in.

Analyzed on private servers. Never used to train models.

FAQ

Questions, answered.

Octym is an AI platform for research integrity and quality from the team behind Proofig: one platform, delivered as tailored solutions for publishers, research institutions, researchers, and integrity & investigation teams, from grant application to final publication. One platform, configured into a solution for the way each team works, with every finding traced to its source.

Octym groups its checks into three categories: Internal manuscript integrity (AI-generated text signals, Image manipulation, Image duplication, AI-generated images, and more); External checks & requirements (Image & graph plagiarism, Retracted reference detection, Hallucinated / non-existent references, Incomplete & broken references, and more); Research quality (Methods & rigor, Reproducibility, Replicability signals, Research ethics, and more). Image integrity and plagiarism are powered by Proofig AI; reference checks are powered by RefGuard, Proofig AI's reference integrity product.

Octym is one platform, delivered as a solution for who you are. Publishers get every journal reviewed its own way, to its own instructions for authors. Research institutions get grant applications checked against funder rules and every manuscript checked against the journals their researchers submit to, before it leaves. Researchers get their work checked against the rules it will actually face. Integrity and investigation teams get a forensic review of any paper, with evidence that holds up. Same platform. A different solution for each.

No. Octym surfaces signals with evidence, not verdicts. Every finding is traced to its source and reviewed by a person, who decides what matters. It does not accuse anyone of misconduct.

Generic AI guesses and can hallucinate. Octym brings every expert lens into one workspace with validated image forensics, evidence traced to its source, an audit trail, and a human in the loop. All tuned to your role.

Yes. Octym includes an AI agent you can talk to in plain language: ask why a finding was flagged, dig into any reference, summarize the reproducibility gaps, or check whether the ethics statement is sufficient. Every answer is grounded in the manuscript and linked to its source.

Yes. Manuscripts are analyzed on secure, private servers and are never used to train AI models. Confidentiality is core to how Octym is built.

For a complete review of a scientific manuscript, Octym is built for the job: it reads the whole paper, not one slice of it. Octym reviews the entire paper across three categories: internal manuscript integrity, external checks and requirements, and research quality. Point tools go deep on one slice, such as figures, prose or citations; Octym reads them all together, with validated image forensics powered by Proofig AI and reference checks powered by RefGuard, in one upload and one report. Every finding is traced to its source, and a person decides.

You upload one manuscript and Octym returns a full review in minutes, not weeks. On a paid package, a submission is one PDF or Word file, up to 50 MB and 100 pages. The review marks every finding on the manuscript, ranks it by severity and traces it to its source, alongside a Summary and a reviewer-style Assessment with major issues at the top. You then confirm, edit or ask Octym's AI agent about any finding.

Both: Octym reads the text and runs image forensics powered by Proofig AI on every figure, inside every review. Octym flags cloned, spliced or retouched images, including Western blots and microscopy; images reused within the paper, even resized, rotated or partly overlapping; AI-generated figures; and duplicated graphs and charts. Figures are also compared against the published literature, powered by Proofig AI, the largest image-plagiarism database in the industry, and legends that describe something other than what is shown are flagged. A person decides what each finding means.

Yes, Octym reads the whole paper in context, so it catches figures, numbers, claims and citations that disagree with each other. Octym flags legends that describe something other than what is shown, citations that do not align with their claims, numbers that differ between text, tables and supplements, and statistics that do not agree. A text-only reader cannot see where a figure contradicts its own legend. Each mismatch is marked where it occurs, for a person to weigh.

Octym reviews the science itself, not only its integrity: research quality carries equal weight alongside the integrity checks. Octym reviews the entire paper across three categories of checks: internal manuscript integrity, external checks and requirements, and research quality. The quality checks cover methods and rigor, reproducibility, replicability signals, research ethics, and novelty and contribution, and a peer-review-grade analysis synthesizes everything into a reviewer-style Assessment, major issues at the top, each comment anchored to its passage. Octym supports a person's judgment, never replaces it.

Yes, Octym reviews grant applications against the funder's own criteria, as well as manuscripts before submission. Review profiles are built for NIH R01, NIH R21, ERC Starting, ERC Consolidator and ERC Advanced, and your Octym team builds any other funder. The review covers scheme fit and eligibility, structure, rigor and preliminary evidence, budget and timeline, data management, and ethics and declarations, and checks that the proposal, CV and budget describe the same project. Each finding is a signal for you to weigh.

Yes, Octym detects hallucinated references: it marks each reference that cannot be matched to a real publication, so invented citations surface early. Its reference checks are powered by RefGuard, Proofig AI's reference integrity product, which compares each reference with the scholarly publishing record to confirm the article is real. It also shows the published details beside each matched reference, so a wrong year or missing author is easy to spot. You then check each entry it could not match and fix or replace it.