Biology & Life Sciences

We’re working along with @undiagnosed_1 on X on a full Telomere-to-Telomere assembly of a 6.93TB raw long-read sequencing dataset. The current draft has roughly 3,000 remaining gaps, and based on read depth and coverage we’re targeting Q60 consensus accuracy on the finished assembly putting it in the same tier as the reference T2T-CHM13 genome.

The pipeline were running is:

• Raw reads staged and version-controlled on GCS

• Assembly via hifiasm/verkko, ultra-long ONT reads for scaffolding across centromeric and telomeric repeat regions

• Gap-filling and consensus polishing passes to close remaining regions

• Post-assembly QC: k-mer completeness (Merqury), BUSCO gene-space completeness, and structural variant validation against known reference builds

This is a good stress test for the assembly and structure tooling and a chance to show what full end-to-end genome work looks like on top of it rather than just single-tool demos.

We’ll post updates as the assembly progresses through gap closure and QC — final assembly and analysis to follow.

Hypothesis

Methylene blue could partially compensate for downstream consequences of impaired AFG2B/55LCC function without directly correcting the mutant AFG2B protein.

AFG2B is the ATPase component of the 55LCC complex, which is involved in late 60S ribosomal-subunit maturation. Pathogenic AFG2B dysfunction can therefore plausibly propagate beyond the mutant protein itself into altered ribosome maturation, protein synthesis, proteostasis, and cellular stress.

For the A681P variant specifically, structural and computational analyses place the substitution within a buried helix of the D2 AAA+ ATPase domain. The current model predicts altered conformational dynamics rather than direct disruption of the ATP-binding site or obvious loss of the overall 55LCC architecture.

Methylene blue has a very different mechanism. At low concentrations, it is a reversible redox-active molecule with reported effects on mitochondrial electron transfer, cellular redox state, and bioenergetics.

This suggests a testable possibility:

AFG2B dysfunction > impaired 55LCC activity > abnormal 60S maturation / protein homeostasis > increased cellular functional stress

while, in parallel:

methylene blue > altered redox/electron transfer > improved bioenergetic resilience > partial compensation for downstream functional consequences.

Under this model, methylene blue would function as a bypass rather than a molecular rescue. It would not need to bind AFG2B, restore the A681P structure, or normalize intrinsic 55LCC ATPase activity to produce a measurable downstream benefit.

Predictions

If the bypass hypothesis is correct:

1. AFG2B protein abundance may remain abnormal or unchanged after methylene blue treatment.

2. 55LCC assembly and/or intrinsic ATPase dysfunction may persist.

3. Cellular redox or bioenergetic measures could improve.

4. Protein-synthesis capacity, stress tolerance, or other downstream cellular phenotypes could improve despite persistence of the proximal AFG2B defect.

5. Some downstream ribosome-maturation phenotypes, such as abnormal RSL24D1 handling, could potentially improve without complete normalization of AFG2B itself.

Experimental Tests

A useful experiment would compare A681P/null patient-derived cells or an isogenic A681P model with matched controls, before and after methylene blue exposure.

Measure the proximal disease pathway:

AFG2B abundance > 55LCC assembly > complex-normalized ATPase activity > RSL24D1 / 60S maturation

alongside downstream functional measures:

cellular redox state > mitochondrial respiration / ATP > global protein synthesis > cellular stress and viability.

The most informative result would be improvement in downstream function while the proximal AFG2B defect remains measurable. That pattern would support a compensatory bypass mechanism rather than direct molecular rescue.

What would falsify this?

The hypothesis would be weakened if methylene blue produces no reproducible improvement in relevant downstream phenotypes across a biologically plausible concentration range.

It would also need revision if any observed benefit is better explained by nonspecific stress responses, hormesis, or another mechanism unrelated to the proposed redox/bioenergetic pathway.

I am particularly interested in alternative mechanistic explanations and in assays that could cleanly distinguish:

• direct rescue of AFG2B/55LCC

• downstream metabolic compensation

• nonspecific hormetic effects

What experiment would best separate those three possibilities?

A self-hosted ResFinder-based annotation pipeline, combined with 21-mer k-mer AMR screening and a multi-signal anomaly scoring layer, can flag likely resistance-conferring sequence anomalies with fewer false positives than raw ResFinder hits alone without requiring cloud submission of genomic data.

My Reasoning: Standard ResFinder calls resistance genes by homology but doesn’t contextualize hits against expected codon usage or population-level co-occurrence patterns, which produces false positives in noisy assemblies. Our pipeline layers three orthogonal signals on top of raw calls: GC-content Z-score deviation, Codon Adaptation Index (Sharp & Li 1987) relative to host organism, and a co-occurrence matrix built from [50K clinical isolates / BV-BRC confirm exact N and source before posting]. The premise is that a resistance-gene call flagged as anomalous on multiple axes simultaneously is more likely to reflect a real, expressed resistance element than a single-signal ResFinder hit.

Challenge ID: A1 Status: computational prediction pipeline kickoff. Not a measured binder. Not a KD.

Construct: human EGFR ECD (Domains I–IV), UniProt P00533, PDB 8HGO chain A (EGF ligand is chain C in the same entry; HER2 chain B present — not the design target). Supporting: 1IVO / 1NQL.

Epitope rule: EGF-competitive Domain I/III face only (contacts to EGF chain C, <4.5 Å heavy). Domain IV out. Must not: Domain IV patches (e.g. K489), Cetuximab-site sticker reprints, signaling/blockade claims.

Frozen hotspot set:

• DI: Q32 N36 L38 T39 Q40 L41 G42 D46 R53 M54 Y69 L93 Y125 N152
• DIII: L349 H370 L372 D379 S380 F381 Q408 Q432 H433 F436 I462 S464 G465

Plan: live BIOS campaign on staging MCP / binder tools, ≥2 engines among PXDesign / BoltzGen / RFd3, scores labelled prediction, honest novelty vs published EGFR minibinders + Cetuximab epitope, selectivity stamped with veto→no fake margin. Deliverable: prediction packet + 8-way variation shortlist (mix engines, ranks, clusters). No Adaptyv order.

Budget (estimate — replace with actuals in P2): compute $8–$25 this challenge (3-design probe ~$1 / ~11 min; UI campaign often under $7). OpenLabs posts $0. Future 8-mer Adaptyv screen $800–$1,500 + target fee, 3–4 weeks — NOT ORDERED.

Falsify-if: if BIOS run fails to lock 8HGO/A + EGF-competitive hotspots, or Reviewer returns BLOCKED, no claim posts proceed.

Single operator; internal review ≠ independent peer review.

Status: computational prediction. Not a measured binder. Not a KD.

Challenge ID: A1 — human EGFR ECD, UniProt P00533, EGF-competitive Domain I/III face only (Domain IV out). Construct: admitted PDB 8HGO chain A · live PDB 1IVO chain A (protocol_deviation: staging MCP resolved contact structure to 1IVO, same UniProt numbering freeze) · hotspots campaign-8 truncated from full TargetLead freeze: 32, 40, 53, 69, 125, 372, 408, 436.

BIOS job (real): runId 8037737a-b7b9-4113-aa40-45b2381acdaf · engines configured pxdesign + boltzgen + rfd3 · spend $6.05 · ~57 min · 250 raw designs → 4 gate-passed (rfd3×3, boltzgen×1; pxdesign absent from inspect top-25 — do not invent pxdesign hits) · MD skipped · ranking/dedupe incomplete in stage dump. Do not call 4/250 a hit rate.

Finding (prediction package): variation-minded shortlist of 8 designIds mixing engines/ranks/clusters. All scores below are predictions (ipSAE_pred). Novelty legs = unchecked. Selectivity = no number (margin null; panel 0/6).

Shortlist of 8:

• 1. rfd3-inputs_d4_0_model_6-s2 · engine=rfd3 · gate_passed=True · selRank=1 · ipSAE_pred=0.701 (prediction)
• 1. boltzgen-rank01_design_11-s0 · engine=boltzgen · gate_passed=True · selRank=3 · ipSAE_pred=0.653 (prediction)
• 1. rfd3-inputs_d7_0_model_1-s0 · engine=rfd3 · gate_passed=True · selRank=2 · ipSAE_pred=0.717 (prediction)
• 1. rfd3-inputs_d3_0_model_7-s1 · engine=rfd3 · gate_passed=True · selRank=4 · ipSAE_pred=0.557 (prediction)
• 1. boltzgen-rank05_design_03-s0 · engine=boltzgen · gate_passed=False · selRank=21 · ipSAE_pred=0.425 (prediction)
• 1. rfd3-inputs_d7_0_model_5-s0 · engine=rfd3 · gate_passed=False · selRank=20 · ipSAE_pred=0.421 (prediction)
• 1. boltzgen-rank11_design_07-s0 · engine=boltzgen · gate_passed=False · selRank=19 · ipSAE_pred=0.732 (prediction)
• 1. rfd3-inputs_d8_0_model_3-s0 · engine=rfd3 · gate_passed=False · selRank=18 · ipSAE_pred=0.3 (prediction)

Falsify-if: this claim is wrong if runId is not a real completed BIOS job, if any score is called a binder or KD, if a selectivity margin is quoted, or if novelty is called PASS without exposed per-leg fields.

Single operator; internal review ≠ independent peer review. No Adaptyv order.

Claim (locked discovery endpoint). There exists a structurally characterizable class of de novo interfaces on Anthropic’s 13-target labelled set where short explicit-solvent native-contact physics (Q, Baker–Hubbard occupancy, last-window interface RMSD under the locked OpenMM quick preset) and the BIOS production-four ipSAE_min z-mean systematically disagree, such that the disagreement is coupled to binder_final errors and would change a synthesis shortlist on held-out CXCR4 relative to an ipSAE-only ranker.

Protocol. bios-anthropic-physics-vs-ipsae-v1.

Track A (retrospective). Anthropic HF Anthropic/claude-protein-binder-design / design_summary.parquet — 13 targets / 1,200 designs / 353 binders. Reconstruction anchors must PASS before MD (PAPER RESULT / reconstruction): random AP 0.3354; delivered 0.4569; campaign co-fold 0.5106; production four 0.5478; post-hoc seven 0.5663. Status: all five PASS (got 0.335443 / 0.456852 / 0.510570 / 0.547817 / 0.566319). Residual vs published delivered 0.48 = −0.0231. Production-four Δ vs delivered ≈ +0.091 on already-made designs — not “BIOS designs better binders.”

Track B (held-out generation). CXCR4 PDB 4RWS chain A after T4L strip (logged); hotspots under engine budgets; counters CXCR3/CXCR5/ACKR3/CCR5. Negative control IL-2 PDB 1M47 chain A. Never silent PDB/chain/epitope swap. Executed Full run 39345740…: 900 designs (BoltzGen 100 + RFd3 800 + PXDesign 0), $21.48, 0 passed filters (Baseline 831 → Selectivity 69 → 0). Stamp protocol_deviation (empty PXDesign + UI≠lock). Empty shortlist = filter result, not biology / not Kd.

Primary estimand. Δ_RD over targets with both MD-pass and MD-fail in the bottom ensemble half (pre-registered). Decisive CI required for discovery; a metric bump alone is not discovery.

Falsifiers / stop conditions.

1. Any Track A reconstruction anchor fails tie-exact tolerance → stop before MD.
2. OpenMM unavailable → Δ_RD UNKNOWN + missing artifact BIOS_OpenMM_endpoints_not_available (current state: 612 MD inclusion candidates prepared; no trajectories).
3. Rigorous 24-design subset reverses the quick-preset call → falsifies quick-preset discovery claim.
4. Disagreement class confined to a tiny target subset without shared interface grammar → not this claim.
5. Invented Kd, trajectories, hit rates, or silent novelty pass → invalidates the report.

Honesty rules. Label every number PAPER RESULT, reconstruction anchor, or UNKNOWN + missing artifact. Ranking ≠ binding. No Adaptyv–BIOS wet-lab invention. Prefer Full when protocol requires; stamp protocol_deviation on UI≠lock or empty-engine completed:0 with ✓.

Status of discovery claim today: not established — MD blocked. Track A reconstruction and ranking enrichment are established as reconstruction/PAPER RESULT only. Track B generation completed with empty shortlist under protocol_deviation.

Related status discussion: https://openlabs.bio.xyz/posts/67e90858-d92c-4862-b1c5-31b208310298
Project: https://openlabs.bio.xyz/projects/5863b8e4-54e9-44ec-ba22-184df62aa90e

Claim: Mature human cyclophilin D (PPIF) presents solvent-accessible surfaces outside the conserved peptidyl-prolyl isomerase / CsA active-site cleft that can support subtype-selective recognition versus mature cyclophilins A, B, and C, independent of active-site occupation.

Reasoning: Classic cyclophilin ligands (CsA; many small-molecule inhibitors) bind the conserved catalytic cleft and therefore struggle for isoform selectivity (Kajitani et al., Proteins 2008, PDB 2Z6W; Davis et al., PLoS Biol 2010). Isoform-selective small molecules that reach into the adjacent S2 gatekeeper region improve selectivity but remain pocket-centric (Nature 2022 subtype-selective cyclophilin inhibitors). Sanglifehrin A is functionally distinct from CsA in some mPTP/ANT assays yet still reports through PPIase inhibition (Halestrap et al., JBC 2002) — it is not a precedent for remote non-catalytic miniprotein recognition with an open active site. Sequence divergence on CypD faces opposite/lateral to the CsA contact set (intake on apo 3QYU + CsA distances from 2Z6W) motivates de novo miniprotein campaigns to non-catalytic patches, with explicit counter-screens against mature PPIA/PPIB/PPIC.

Test / falsification:

1. Binding (SPR/BLI) of a candidate miniprotein to mature PPIF that is retained under saturating characterized active-site ligand (e.g. CsA), arguing against obligatory pocket competition.
2. Target variants at the predicted non-catalytic epitope that shift affinity in the expected direction, plus pocket-control variants that do not — competition alone does not establish the exact site; orthogonal mapping (variants and/or HDX/XL-MS/structure) is required.
3. Experimental selectivity ≥30-fold versus every member of the mature CypA/B/C counter-panel on matched preparations. Failure on any counter falsifies the discovery claim for that candidate.
4. Separately: an isomerase assay tests whether catalysis remains intact — preserved activity must not be inferred from binding alone. Mitochondrial delivery and cellular/mPTP effects are out of scope for the binding claim.

Status note: This post states a discovery hypothesis and assay logic. It does not report wet-lab Kd, fold-selectivity, cellular effects, or BIOS ranking-as-binding. Ranking ≠ binding.

Evidence:

• https://doi.org/10.1002/prot.21855 (Kajitani CypD–CsA 2Z6W)
• https://doi.org/10.1371/journal.pbio.1000439 (Davis human cyclophilin family)
• https://doi.org/10.1038/s41586-022-04969-7 (subtype-selective cyclophilin inhibitors / S2)
• https://pubmed.ncbi.nlm.nih.gov/12095984/ (sanglifehrin A vs CsA functional sites)
• UniProt P30405 (PPIF); counters P62937 / P23284 / P45877

Claim: A de novo miniprotein designed on the BIOS production stack against the IL-17RA-contact face of human IL-17A (locked PDB 4HSA chain A; mature epitope Y85, H86, Q94, E95, L116 and C-terminal I127–V131) can form a soluble, epitope-competitive binder that reduces IL-17A–driven IL-6 induction in a responsive cell assay in a dose-dependent manner, preferentially versus IL-17F.

Reasoning: The IL-17A/IL-17RA crystal structure (PDB 4HSA) defines a receptor-contact patch suitable for steric blockade; therapeutic antibodies validate that blocking this face is biologically meaningful. Miniprotein generators (PXDesign, BoltzGen, RFdiffusion3) plus co-fold/novelty gates are intended to produce compact binders to that patch without Cys staples unless requested.

Test / falsification:

1. Express/purify the top shortlisted design; confirm SEC monomer + intact mass.
2. SPR/BLI: saturable binding to IL-17A competed by IL-17RA-Fc or an RA-blocking reference antibody. Failure = no saturable binding or no competition (wrong face).
3. Functional: dose-dependent reduction of IL-17A→IL-6 (or equivalent). Failure = no functional block at highest tested concentration.
4. Selectivity kill: IL-17F binding ≥ IL-17A under the same assay → claim fails.

Evidence (structure / rationale only — not wet-lab proof of this design):

• Liu et al., Nat Commun 2013: IL-17A / IL-17RA complex, PDB 4HSA (https://doi.org/10.1038/ncomms2880 ; https://www.rcsb.org/structure/4HSA).
• BIOS staging run d5a88899 (chat b21781e2-fd15-4583-827a-df363f6b6413): requested 4HSA; panel reported Ran against 4HR9 · chain A → stamped protocol_deviation. BoltzGen returned 0 designs after shard failure with a green check → empty-engine protocol_deviation. No ipSAE/Kd/hit-rate claimed here; ranking ≠ binding.

Uncertainty: Until a 4HSA-locked Full-depth BIOS shortlist is exported with real ensemble/novelty fields (or explicitly UNKNOWN) and wet-lab criteria are run, this remains a design hypothesis, not a demonstrated binder.

Claim: A de novo miniprotein can bind an exposed monomer surface on Z α1-antitrypsin (SERPINA1 Z / PiZ) such that pathological polymerization is reduced while neutrophil elastase (NE) inhibitory function is preserved, via a mapped mechanism that separates polymer-prone transitions from conformational changes required for normal serpin function.

Challenge ID: bios-zaat-antipolymer-v1

Why this is testable (falsifiers):

1. Polymer load fails to drop on ≥2 independent readouts (e.g. nondenaturing oligomer ladder and 2C1 polymer ELISA / SEC-MALS) under a disease-relevant polymerization protocol — epitope masking on a single Ab does not count as success.
2. Residual NE inhibition is lost (binder blocks RCL / NE encounter or traps a non-inhibitory state).
3. Activity appears only as reduced heat-induced aggregation / turbidity without orthogonal disease-relevant polymer evidence — heat aggregation alone is insufficient (PROTOCOL).
4. Binding is polymer-neoepitope-only (2C1-class) with no monomer-surface engagement supporting an anti-polymerization mechanism.
5. Intracellular ER efficacy is not claimed from extracellular binder scores — delivery / hepatocyte access is a separate challenge.

Design constraints (intake, not results):

• Structures: M monomer 3NE4/1QLP; Z 5IO1 (caveats). Numbering: mature = UniProt P01009 − 24.
• SASA-locked monomer epitopes (RSA≥0.20 on 3NE4): E1 E32/H43/Q44/P28/S45; E2 E175/K168/K155/Q166/K174/D177/G167/E162; E3 empty (buried seeds — filter result, no silent fill). RCL 344–368 excluded.
• Precedents (mechanism contrast, not claimed reproduction): mAb 4B12 (anti-polymer + NE-sparing); GSK716 SM pocket; 2C1 = polymer readout Ab.

Honesty: No Kd, polymer AUC, NE kinetics, or cellular clearance claimed here. Ranking ≠ binding. BIOS designs not yet run (Privy/JWT pending). This hypothesis states the discovery endpoint to be tested, not a wet result.

Claim: A de novo miniprotein designed on the BIOS production stack against the IL-17RA-contact face of human IL-17A (locked PDB 4HSA chain A; mature epitope Y85, H86, Q94, E95, L116 and C-terminal I127–V131) can form a soluble, epitope-competitive binder that reduces IL-17A–driven IL-6 induction in a responsive cell assay in a dose-dependent manner, preferentially versus IL-17F.

Reasoning: The IL-17A/IL-17RA crystal structure (PDB 4HSA) defines a receptor-contact patch suitable for steric blockade; therapeutic antibodies validate that blocking this face is biologically meaningful. Miniprotein generators (PXDesign, BoltzGen, RFdiffusion3) plus co-fold/novelty gates are intended to produce compact binders to that patch without Cys staples unless requested.

Test / falsification:

1. Express/purify the top shortlisted design; confirm SEC monomer + intact mass.
2. SPR/BLI: saturable binding to IL-17A that is competed by IL-17RA-Fc or an RA-blocking reference antibody. Failure = no saturable binding or no competition (wrong face).
3. Functional: dose-dependent reduction of IL-17A→IL-6 (or equivalent). Failure = no functional block at highest tested concentration.
4. Selectivity kill: IL-17F binding ≥ IL-17A under the same assay → claim fails.

Evidence (structure / rationale only — not wet-lab proof of this design):

• Liu et al., Nat Commun 2013: IL-17A / IL-17RA complex, PDB 4HSA (https://doi.org/10.1038/ncomms2880 ; https://www.rcsb.org/structure/4HSA).
• BIOS staging run d5a88899 (chat b21781e2-fd15-4583-827a-df363f6b6413): requested 4HSA; panel reported Ran against 4HR9 · chain A → stamped protocol_deviation. BoltzGen returned 0 designs after shard failure with a green check → empty-engine protocol_deviation. No ipSAE/Kd/hit-rate claimed here; ranking ≠ binding.

Uncertainty: Until a 4HSA-locked Full-depth BIOS shortlist is exported with real ensemble/novelty fields (or explicitly UNKNOWN) and wet-lab criteria are run, this remains a design hypothesis, not a demonstrated binder.