DarkBiome
Configure & order
Computational Biology & Biophysics Services

Catch the broken ones first.

Nine fixed-fee computational services across three R&D stages, run on the same machinery we use for our own discovery programs. Send candidate sequences, draft patent claims, database accessions, long reads, or 96-well assay recordings — get back a plain verdict table, the numbers behind it, and every raw file.

For enzyme, binder, biologic, and biomaterial teams across bacterial, yeast, mammalian, plant, filamentous, and cell-free expression with more candidates than they can afford to build — and no computational biophysicist on staff.

Fixed fee, agreed up front One project at a time Local workstation by default

Forty candidates, six synthesized Forty squares in a grid. Six are filled, standing for the candidates a typical synthesis budget covers; the remaining thirty-four are empty outlines. A SCREEN RETURNS 40 6 fit the synthesis budget

A screen returns forty candidates; the synthesis budget covers six. Pick those six off a spreadsheet of sequence properties, and the ones that were never going to fold cost you four weeks and a synthesis slot each. Simulating them first takes days, and catches the ones that come apart before you order.

Which one do you need?

Track 1 • Screen & Order

More candidates than synthesis budget

Check boundaries, host ER/glycosylation traps, ligand strain, and water/heat stability before ordering DNA.

$5,000–$10,0003–5 days
Service 01 → Track 1 • Screen & Order

A working lead with stability or PTM liabilities

Scrub unpaired cysteines, glycan shields, deamidation, protease, and aggregation risks while locking the 3D active site.

$8,000–$15,0005–7 days
Service 08 → Track 1 • Screen & Order

A repeat gene your vendor rejected

Redesign the DNA for the exact same protein so vendors accept it and your host keeps repeats intact.

$4,000 on acceptance3–5 days
Service 02 → Track 2 • Find & Rescue

A database hit that looks cut off or suspect

Rebuild the full gene from raw reads, check co-operon helpers and contig seams, and verify the true source organism.

$3,500–$7,5005 days
Service 05 → Track 2 • Find & Rescue

Need an alternative to a patented protein

Find low-identity natural relatives by 3D shape, reconstruct heat-stable ancestors, and compress into a diverse plate.

$6,000–$14,0007–10 days
Service 06 → Track 2 • Find & Rescue

Need compact peptides, knottins, or sORF multimers

Mine cleaved prohormone cassettes, graft 6-Cys knottin loops, or find short SPPS/AAV homo-oligomers.

$10,000–$18,0007–10 days
Service 09 → Track 3 • Verify & Audit

Engineered clones you have not verified

Find where the construct landed, how many copies integrated, and whether tandem arrays collapsed.

$2,500 (up to 4 clones)5 days
Service 03 → Track 3 • Verify & Audit

Freezing videos or plate assays, unanalyzed

Turn any thermal camera video — or overhead video + thermocouple log — into calibrated spectra.

$1,500/recording3 days
Service 04 → Track 3 • Verify & Audit

Filing or licensing a sequence patent

Check draft claims and ST.26 XML listings for math contradictions, residue-offset errors, and public sequence/3D collisions.

$4,500–$9,5002–3 days
Service 07 →
Compare all 9 services side by side — inputs, deliverables, turnaround & fixed fees
Service Track You send Core deliverable Turnaround Fixed fee
01 • Stability screen 1 · Screen & Order 20–50 FASTA or PDB candidates Holds/Marginal/Fails table, host/glycan/PTM/SPPS flags, 300 K/330 K MD metrics & PDBs 3–5 days $5,000–$10,000
08 • Protein optimization 1 · Screen & Order 1–3 working parent sequences/PDBs 10–25 active-site-locked, liability-scrubbed, 3D-refolded & water-tested variants 5–7 days $8,000–$15,000
02 • Synthesis rescue 1 · Screen & Order Repeat protein AA/DNA + target host Vendor-accepted repeat-scrubbed DNA + multi-copy siblings ($750 pre-check option) 3–5 days $4,000 on acceptance
05 • Broken gene rescue 2 · Find & Rescue 1–10 database hits + SRA/raw reads Full-length repaired gene, co-operon/chimera check, 3D PDBs & true organism verdict 5 days $3,500–$7,500
06 • Homolog & ASR design 2 · Find & Rescue Reference sequence/PDB + patent refs Diverse plate of <25%-id 3D homologs + ancestral designs + identity/3D tables 7–10 days $6,000–$14,000
09 • Peptide & miniproteins 2 · Find & Rescue Target motif/metal or source genomes Cleaved cryptic peptides, 6-Cys knottin grafts & sORF multimers + SPPS scorecard 7–10 days $10,000–$18,000
03 • Clone audit 3 · Verify & Audit Nanopore/HiFi reads + host/construct On/off-target integration map, copy number & raw-read tandem array collapse report 5 days $2,500 (up to 4 clones)
04 • Assay analysis 3 · Verify & Audit 96-well thermal or optical video + thermocouple CSV Daybreak f(T), T50, Vali K(T), Koop Δa_w, uncertainty budget, CNT footprint & NetCDF 3 days $1,500/recording
07 • Patent & ST.26 audit 3 · Verify & Audit Draft claims/spec + ST.26 XML + PDBs Claim math check, public sequence/3D collision table, residue/codon map & ST.26 log 2–3 days $4,500–$9,500
Track 1 · Screen, Optimize & Order DNA

Cut synthesis waste, fix sequence liabilities & pass vendor DNA intake

Use these three services when you already have candidate sequences or AI designs and want to drop the ones that unfold or misfire in your expression host (01), surgically scrub unpaired cysteines, chemical degradation sites, or aggregation patches without touching your active site (08), or get a rejected repeat gene past vendor synthesis screens (02).

SERVICE 01 • TRACK 1

Pre-synthesis stability screen

Before you spend months expressing a sequence, we test whether the full protein holds its fold in water and heat and check that its start and end boundaries aren’t cut off. We also flag free cysteines, active-site glycosylation traps, chemical degradation motifs, strained ligand poses, and peptide synthesis stalls across your chosen host (E. coli, Bacillus, yeast, mammalian, plant, filamentous, or cell-free). Built for batches of 20–50 enzymes, binders, repeat proteins, short peptides, giant proteins up to 2,000 residues, or multi-chain oligomers — whether natural database hits or designs from AlphaFold, ESMFold, Boltz-2, OpenFold 3, RFdiffusion, ProteinMPNN, or BindCraft. Every candidate runs in water and salt at 300 K (27 °C), then steps to 330 K (57 °C) in OpenMM to see whether its fold, active site, subunit interface, and metal cage stay intact.

When you should not buy this

If your question is binding affinity, measure it rather than predict it. For small, well-behaved binders against a single known target, high-throughput services return a measured KD at a price no prediction can honestly compete with, and published benchmarks show computational binding metrics correlating only weakly with the measured value. We would rather say so than take the project. What those services do not cover is the list above: a construct that will not express because of cut-off ends, an active-site glycan shield, or a host disulfide mismatch, a multi-chain assembly, a metal-loaded site, or whether a fold survives heat rather than whether it binds. That is where host checks and water simulation earn their place, and it is the only place we will sell it to you.

Tiers, protocol and what you receive
Step 1 • Boundary & giant-fold QC  Catches cut-off N- or C-terminal ends, missing structural end-caps, and miscalled start sites; folds giant proteins (410–2,000 AA) via repeat-phase-matched overlapping windows with SVD stitching, topological knot rejection, and seam hydrogen-bond checks
Step 2 • Host, glycan-shield, PTM & SPPS check  Separates real 3D disulfide bonds from unpaired cysteines that stall ER secretion; flags N-linked (N-X-S/T) and dense O-linked glycosylation sites positioned where host glycans block the active site or binding face in yeast, fungal, mammalian, or plant hosts; scans 9 chemical and protease liabilities (NG/NS/NT deamidation, DG isomerization, DP acid cleavage, exposed Met oxidation, dibasic protease clips, MHC-II 9-mer cores); and scores short peptides (≤85 AA) for solid-phase synthesis (SPPS) coupling stalls and Native Chemical Ligation (NCL) junctions
Step 3 • Ligand strain, multimer & water/heat MD  Checks co-folded substrates/cofactors for pose agreement and high-energy internal conformational strain; verifies hetero-oligomer cross-interface co-evolutionary contact support and drops homo-oligomers that pack as catalytically dead “inert bricks” or inverted micelles; then runs all-atom OpenMM simulation in water and 150 mM NaCl (27 °C → 57 °C), tracking backbone drift, dynamic catalytic-triad closure, and polarizable divalent/trivalent metal-cage retention (12-6-4 model)
TierPer candidateThe question it answersTypical use
Triage 0.5 ns Does the fold survive contact with solvent and a thermal ramp at all? A whole screen, 20–50 candidates
Standard 10 ns Once the fast failures are gone, does it still hold its shape and its interface? A shortlist of about ten
Extended 50 ns+ ×3 Does the answer reproduce across independent runs, or did one trajectory get lucky? The 3–5 you are about to commit budget to

Extended runs are triplicate with independent starting velocities, and we report the spread across replicas rather than a single mean, so you can judge how firm the result is. Most projects run Triage across the full set, then take the survivors up a tier.

You send

  • 20–50 sequences in FASTA (monomers, multimers, short peptides, or giant proteins up to 2,000 AA)
  • Chosen synthesis or expression route (SPPS/NCL, E. coli, yeast, mammalian, plant, filamentous, or cell-free)
  • Any 3D structures you already have
  • Assembly stoichiometry, active-site residues, substrates/cofactors, or bound metals, if applicable

We check & measure

  • Intact start/end boundaries, structural end-caps & knot-free giant chunk stitching
  • 3D disulfides vs. unpaired ER cysteines, active-site glycan shielding, 9 chemical/protease/MHC-II liabilities & SPPS coupling stalls
  • Backbone drift and unfolding in water at 27 °C and 57 °C
  • Ligand pose strain, multimer “inert brick” rejection, dynamic triad persistence & polarizable metal-cage retention

You receive

  • A clear table of every candidate: host/glycan/PTM/SPPS flags, stability numbers, and pass/marginal/fail band
  • One summary page per candidate
  • Every 3D model as PDB
  • All metrics as JSON, ready to plot
  • A plain methods section you can paste into a report

Each candidate lands in one of three bands. The cutoffs are fixed before your run, never chosen afterward to suit the results, and they are set from matched intact and destabilized control structures run alongside your candidates at no extra charge (plus any reference controls you supply) rather than read off a fixed table:

Holds    keeps its fold, and its interface, through the 330 K stress step
Marginal  moves more than a well-behaved control, but does not come apart
Fails     loses its fold or its interface under the 330 K stress step

This page used to print fixed numeric cutoffs. We took them down because our own benchmark showed they were too permissive: control structures we had deliberately broken, with most of the chain cut away and the core opened to solvent, still cleared the failing threshold. A band nothing can land in is not a verdict. Cutoffs go back here as numbers once they are confirmed on structures that were not used to derive them; until then you get the exact cutoffs applied to your run, and the control results behind them, in the report.

Our benchmark is written down, including the parts that went badly. We tested whether a 0.5 ns 300 K → 330 K run can rank five matched thermophile-versus-mesophile natural homolog pairs by stability: it scored 2 of 5 correct (one-sided sign test p = 0.812), no better than a coin flip, so we do not claim it ranks surviving natural proteins. What it did do in our pre-registered intact-versus-broken control benchmark (ten intact crystallographic structures against five excised central-40% internal-fragment negative controls): 10 of 10 intact folds held (mean 330 K core Cα RMSD 1.13 Å, range 0.61–1.86 Å), whereas the destabilized internal-fragment controls unpacked to a mean 330 K Cα RMSD of 2.49 Å (range 1.56–3.18 Å) — a 1.36 Å mean separation — while also confirming intact four-cysteine Fe–S4 coordination in Pyrococcus furiosus rubredoxin (1BRF, min Fe–S 2.13 Å). On the host-secretion and quaternary side, secreted proteins routinely carry 3D-unpaired cysteines that stall eukaryotic ER export alongside true structural disulfide bonds that naive cysteine removal breaks — and over a third of homo-oligomer designs clearing high static interface confidence pack tightly with zero closed active sites or collapse as inverted micelles. Static confidence scores pass both failure modes. We also threw away our first simulation attempt after finding our own sample-preparation artifacts inside it (a two-chain asymmetric unit in 1C9O and 348 riding crystallographic hydrogens in 5RXN). You can inspect the full benchmark table and sample deliverable previews below before you commit a budget.

Fee $5,000 – $10,000
Turnaround 3–5 business days
Capacity Up to 50 sequences

The 3–5 day figure is Triage across a full screen: three business days for up to 24 sequences, five for 25–50, measured on our own hardware rather than estimated. Standard and Extended tiers run longer and we quote their schedule in writing at intake, because it depends on sequence length, oligomeric state and how much GPU capacity you want us to add. Fee scales with tier, sequence count, and whether multimer interfaces, ligand poses, or metal centers are involved.

What this does not tell you

No tier here measures thermodynamic stability. Triage is a short-timescale filter: it separates folds that come apart quickly under a thermal ramp from folds that hold their shape. It catches candidates that were never going to work; it is not evidence that a survivor is stable. Standard and Extended make that same call more confidently, not a different call. They cannot put two survivors in order: we tested exactly that, and it scored no better than chance. None of it predicts expression yield, solubility in your buffer, aggregation over days, or catalytic activity. Use it to drop what comes apart, not in place of an assay.

Scope & order this service →
SERVICE 08 • TRACK 1

Constraint-guided in silico protein optimization & liability scrubbing

Your enzyme, binder, repeat scaffold, or homo-oligomer works on the bench, but hits trouble in manufacturing or storage: free cysteines stall eukaryotic secretion, glycans block substrate entry, deamidation or protease sites shorten shelf life, or exposed patches trigger aggregation and immune recognition. Unconstrained AI redesign often fixes those problems by breaking the active site. We freeze your catalytic residues, cofactor pocket, and subunit interfaces in 3D while mutating only permissive scaffold positions — returning 10–25 refolded, water-simulated variants ready to test as a single synthesis plate.

When you should not buy this

If your wild-type sequence expresses cleanly, stores well, and meets your process target today, do not touch it — every unnecessary mutation carries epistatic risk. And if your starting protein has zero measurable activity and you need a brand-new catalytic mechanism invented from scratch, constrained scaffold optimization will not create chemistry that isn’t there. Buy this when the core function works (or comes from a validated family) and specific expression, redox, glycosylation, chemical-stability, immunogenicity, or aggregation liabilities stand between your lead and scale-up.

How it works and what you receive
3D coordinate lock  Freezes active-site triads, oxyanion holes, metal/cofactor shells, true 3D structural disulfides, and symmetric homo-oligomer packing contacts across C2–C6 or D2 interfaces
Multi-constraint optimization  Removes 3D-unpaired ER cysteines, active-site N- and O-linked glycosylation traps, and 9 chemical/protease/MHC-II liabilities (NG/NS/NT deamidation, DG isomerization, DP acid cleavage, exposed Met, dibasic protease clips, MHC-II 9-mer cores) while preserving surface solubility, blocking amyloid-prone patches, matching target pI/charge, and keeping repeat axes straight
Refold & water check  Independently re-folds every evolved sequence in 3D (requiring high backbone TM-score to your parent fold and closed catalytic geometry) and runs our 300 K → 330 K OpenMM explicit-water stress check on the delivered panel

You send

  • 1–3 parent protein sequences or 3D structures (monomer, repeat protein, or homo-oligomer)
  • Residues or regions that must remain untouched (active site, binding epitope, tags)
  • Target expression host, formulation pH/salt, and liabilities to eliminate

We optimize

  • 3D-unpaired cysteine removal without breaking real structural disulfides
  • Active-site glycan-shield removal, chemical PTM (NG/DG/DP/Met), host protease & MHC-II core scrubbing
  • Surface solubility, aggregation-patch removal & repeat-scaffold straightening
  • Symmetric homo-oligomer interface & axial pore electrostatics

You receive

  • 10–25 variant sequences (from conservative single/double edits to broader surface cleanup)
  • Before/after liability table, mutation rationale & 3D PDB models
  • 300 K / 330 K OpenMM explicit-water stability metrics against your wild-type control
Fee $8,000 – $15,000
Turnaround 5–7 business days
Deliverable 10–25 verified 3D variants

Fee depends on scaffold length, monomer vs. symmetric multimer optimization, and simulation tier on the final variant panel. Repeat-free host DNA design (Service 02) can be bundled for immediate gene ordering.

What this does not tell you

Removing sequence liabilities and confirming that the 3D backbone and active site hold in water reduces the fraction of dead designs on your plate, but distal mutations can still alter catalytic rate (kcat), substrate affinity (KM), dynamic loop gating, or host expression titer in ways current force fields and sequence models miss. MHC-II 9-mer core removal is a sequence-level de-immunization filter, not a clinical T-cell assay. Always assay the delivered 10–25 variants alongside your unmodified parent control.

Scope & order this service →
SERVICE 02 • TRACK 1

Repetitive gene design & synthesis rescue

When a gene vendor rejects your repeat protein — or your host keeps deleting repeats — we redesign the DNA sequence encoding the exact same protein so vendors accept the order and the gene stays stable across bacterial, yeast, mammalian, plant, filamentous, or cell-free expression. Built for any repeat or low-complexity protein: collagens, silk/elastin/resilin fibers, repeat binders, solenoids, CRISPR/TAL arrays, and mucins. Free vendor codon tools reuse the same preferred codon on every repeat (causing vendor rejection and DNA loop-out in the host), while random scrambling inserts slow codons that stall protein translation. We vary synonymous codons cleanly across every repeat without using slow codons, remove hidden splice sites, mRNA hairpins, and cloning-enzyme sites, and keep your amino acid sequence 100% unchanged.

When you should not buy this

If you are ordering a single non-repetitive protein and your vendor’s free codon tool accepts it, use the free tool. You do not need us. This service is built for the cases where vendor portals fail: repeat proteins rejected at order intake, plasmids that lose repeats during cloning in E. coli, and multi-copy yeast, mammalian, plant, or fungal strains where identical DNA copies recombine and delete themselves during growth.

Constraints and what you receive
Repeat removal  Breaks up exact DNA repeats across the entire gene (strict enough for standard synthesis at Twist, IDT, GenScript, or enzymatic synthesis) • distinct DNA versions for multi-copy strains so copies cannot recombine • balanced GC (35–65% overall, 25–70% in every 50 bp window)
Clean expression  Bans slow rare codons • clears 5′ mRNA hairpins around the start codon • strips hidden splice sites, early stop/polyA signals, and homopolymer base runs • optional host-native regulatory elements and 3D-disulfide-preserving ER-thiol scrubbing (Service 08 add-on) for eukaryotic secretion
Ready to clone  Clears 11 Golden Gate & standard restriction enzymes (BsaI, BbsI, BsmBI, SapI, AarI, EcoRI, BamHI, HindIII, XhoI, NdeI, NotI) on both strands and across part junctions
Hosts  17 built-in host profiles across mammalian, insect, yeast, filamentous fungi & bacteria • any other organism on request

Why vendor tools fail

Free vendor portals pick the same top codon on every repeat unit, triggering their own repeat-complexity rejection and causing DNA deletions in cloning and production strains. Naive repeat-breakers fix the DNA repeats by inserting rare codons that stall translation, and ignore mRNA folding at the start site, hidden eukaryotic splice sites, and restriction sites formed at part junctions.

What we redesign

We spread synonymous codons across fast host tRNAs so no long DNA repeat survives, keep translation fast, clear start-site hairpins and unwanted splice or termination signals for your expression host, clear cloning-enzyme sites on both strands, and generate distinct DNA copies when you want multiple copies in one genome.

You receive

  • Synthesis-ready DNA sequence(s) + optional distinct sibling copies for multi-copy strains
  • Cloning-ready expression cassette (promoters, secretion tags, optional host regulatory elements, homology arms), junction-checked
  • A plain QC report: repeat check, GC profile, codon fit, mRNA & restriction audit
  • Protein translation verified 100% identical to your input
Expression hosts
GroupOrganisms
Mammalian & insect CHO, HEK293, Sf9
Yeast Saccharomyces cerevisiae, Pichia pastoris (Komagataella phaffii), Yarrowia lipolytica
Filamentous fungi Aspergillus niger, Aspergillus oryzae, Trichoderma reesei
Bacteria Escherichia coli, Bacillus subtilis, Bacillus licheniformis, Bacillus megaterium, Corynebacterium glutamicum, Pseudomonas putida, Lactococcus lactis, Vibrio natriegens
Plant, cell-free & other Any other organism on request, including plant expression and bacterial or eukaryotic cell-free systems.
Fee $4,000 per construct
Payable Only once your vendor accepts
Turnaround 3–5 business days

We carry the risk, because that is where the value is: a complexity surcharge costs a few hundred dollars and buys an accepted sequence, but a refusal cannot be paid past at any price, and the construct stays unbuildable. No deposit is due on signature. Vendor acceptance means the redesigned sequence clears your target vendor's automated sequence-complexity and order-intake screen for standard synthesis (verified on the vendor portal or within ten business days of delivery unless you share a vendor refusal notice). If the redesign is still refused at order intake, there is no fee.

Before you submit • $750 • 2 business days

A pre-submission check, if you would rather find out now than after a two-week queue. We score your construct against each vendor's published acceptance criteria and report the longest exact repeat, repeat content at that vendor's threshold, inverted repeats, windowed GC, homopolymer runs and restriction sites, with a verdict per vendor. Credited in full against a redesign if you go ahead.

What this does not tell you

This is a design tool. Its output has not been validated at the bench, by us or by anyone else, and we make no claim about expression level or in vivo cassette stability. Acceptance criteria differ between vendors and most do not publish them at all: we design and check against the criteria that are published, and for the rest we work from the refusal you received. Criteria also change without notice, which is why the fee is contingent rather than guaranteed. We would rather carry that uncertainty than describe it as solved.

Scope & order this service →

Designing a repeat construct against recombination and confirming it survived integration are the same problem at two points in time. If you are heading for a genomic cassette, Service 03 checks what actually landed.

Track 2 · Find & Rescue Sequences

Repair broken database entries, find low-identity 3D homologs & mine compact peptides

Use these three services when you need better starting sequences: rebuilding truncated or mislabeled database hits and checking for missing co-operon helper genes (05), finding distant (<25% identity) natural homologs and heat-stable ancestral proteins outside crowded patent estates (06), or discovering cleaved peptides, 6-Cys knottins, and short sORF homo-oligomers sized for SPPS or single-AAV vectors (09).

SERVICE 05 • TRACK 2

Broken gene, operon & contamination rescue

Public genome and metagenome databases are full of cut-off sequences, frameshifts, missed exons, chimeric assembly seams, missing co-operon helper genes, low-complexity masking blind spots, and mislabeled microbial contaminants. When a promising database hit fails to express, the biology is often fine — the database entry was simply cut in half at an assembly break, started at the wrong methionine, left its required folding chaperone behind in the genomic locus, or came from a hitchhiking bacterium mislabeled as a plant, marine, or eukaryotic gene. Send us a database hit that looks incomplete or failed in the lab: we rebuild the full, intact gene from raw sequencing reads, audit the surrounding operon, check the complete 3D protein fold, and confirm which organism it actually came from before you order DNA.

When you should not buy this

If your sequence comes from a well-studied reference organism backed by full-length mRNA data or an experimental crystal structure, with clean start and end boundaries, there is nothing here to rescue — go straight to synthesis or Service 01. Buy this when your candidates come from environmental metagenomes, draft genome assemblies, or unverified plant, marine, fungal, or microbial database entries that have never been checked against the underlying sequencing reads.

What we check and what you receive
Gene reconstruction  Pulls the original raw sequencing reads (with low-complexity repeat masking switched off so repetitive exons and split contigs aren’t silently dropped) to bridge assembly breaks and recover missing N- and C-terminal ends
Gene-model & chimera repair  Fixes miscalled start codons, missed short exons, and sequencing frameshifts, and flags chimeric metagenomic mis-assemblies (“Frankenstein” contigs stitched from two different organisms across a k-mer repeat)
Operon helper & 3D fold check  Inspects flanking genes (±10 kb) for required folding chaperones, maturation proteases, cofactor-loading proteins, or partner subunits that were left behind when only the catalytic ORF was ordered, and folds the repaired full-length protein in 3D alongside the raw database fragment
Source check  Checks neighboring genes, intron structure, DNA composition, read depth, and evolutionary relatives to confirm whether the gene truly belongs to the listed organism or a microbial contaminant

You send

  • 1–10 database accessions or candidate sequences (any enzyme, binder, biomaterial, or biologic)
  • Links to public sequencing reads (NCBI/ENA) or your own raw FASTQ/assembly files
  • Your target expression host (E. coli, Bacillus, yeast, mammalian, plant, filamentous, or cell-free)

We check & rebuild

  • Sequences cut off at contig edges, masked out by repeat filters, or split across two entries
  • Wrong start sites, missed secretion signals, skipped exons, frameshifts & chimeric contig seams
  • Missing co-operon chaperones, maturation proteases, or partner subunits in the native locus
  • True organism of origin vs. hitchhiking bacterial or fungal contamination

You receive

  • Repaired full-length DNA and protein FASTA sequences (plus any required co-operon helper sequences)
  • Read-support summary, repair log & before/after 3D PDB models
  • Source-organism verdict + recommended expression host and compartment

Why this check exists: across draft NCBI and JGI assemblies, repetitive and multi-domain genes are routinely broken by four recurring database artifacts — short-read contig breaks inside repeats, local alignment clipping that drops N- and C-terminal structural caps, gene callers splitting one protein into two separate entries, and internal methionine mis-calls that delete native secretion signals — while high-profile “eukaryotic” database hits in marine, plant, and algal assemblies frequently turn out to be mislabeled hitchhiking bacterial contigs lacking introns and eukaryotic flanking genes.

Fee $3,500 – $7,500
Turnaround 5 business days
Scope 1–10 database hits

Fee scales within the range by the number of candidates and whether raw sequencing reads need to be re-assembled versus checking existing genome contigs and flanking operons.

What this does not tell you

Rebuilding a cut-off gene requires raw sequencing reads or overlapping assemblies that actually cover the missing piece; if the original sequencing run never captured that region, we report the exact gap rather than guessing missing amino acids. Where a frameshift appears in the raw reads, we separate sequencing errors from naturally inactivated pseudogenes and label them clearly. Repairing a broken sequence prevents ordering half a protein or expressing a bacterial contaminant in the wrong host, but it does not measure enzyme activity or yield.

Scope & order this service →
SERVICE 06 • TRACK 2

Patent-clear homolog search & ancestral protein design

Need an alternative to a patented enzyme, binder, or structural protein, or a more heat-stable starting scaffold without rounds of wet-lab evolution? Standard sequence search goes blind below ~25% identity, and default low-complexity filters erase repetitive homologs even above 50% identity — keeping R&D teams trapped in the same crowded sequence space as existing patents. Because 3D active-site shape stays conserved long after sequence diverges, we search public genomes and metagenomes by 3D structure with repeat masking switched off, reconstruct heat-stable ancestral proteins from the family tree, and compress thousands of hits into a diverse 24-, 48-, or 96-well synthesis plate.

When you should not buy this

If an unencumbered public protein at 60% sequence identity already works at your temperature and pH, a five-minute BLAST search will find it — do not pay us for deep 3D search. And if what you need is a formal legal freedom-to-operate opinion or claim chart sign-off, hire registered patent counsel: we deliver the low-identity candidate sequences, 3D structural comparisons, and sequence-identity tables that your scientists and patent counsel review, not legal advice.

How it works and what you receive
Unmasked + 3D search  Finds distant natural relatives (<25% sequence identity, plus repetitive homologs hidden by default BLAST/DIAMOND low-complexity masking) across public genomes and environmental metagenomes, using composition-preserving shuffle controls to cut biased-composition false positives from 98.0% to 0.13%
3D structural tree & ancestral design  Builds an all-vs-all 3D structural and sequence family tree across diverse species (separating true functional active-site geometries from collapsed steric-zipper or backbone-only decoys), reconstructs ancestral protein sequences from internal nodes, and checks every candidate in 3D for clean folding, intact catalytic geometry, and host ER-redox compatibility
Plate diversity compression  Clusters thousands of raw hits by 3D active-site geometry, sequence distance, taxonomy, and pI/solubility so your 24-, 48-, or 96-well synthesis plate samples every distinct clade instead of wasting 80% of your budget on near-duplicate sequences

You send

  • Reference protein sequence(s) or 3D structure(s) and key active-site or binding residues
  • Patent or competitor sequences to measure percent sequence identity against (optional)
  • Target plate size (24, 48, or 96 wells), operating conditions (temperature, pH, cofactors) & expression host

We deliver

  • Distant natural relatives (<25% identity) matched on 3D active-site geometry
  • Reconstructed ancestral protein designs built for higher heat stability
  • Diversity-compressed synthesis plate panel with 3D folding and host-compatibility filtering

You receive

  • Synthesis-ready panel of distant natural homologs & ancestral designs (FASTA + 3D PDB models)
  • Full-length, domain-level, and 3D structural similarity tables against every reference sequence you supply
  • Family tree files, structural alignments, confidence scores & host expression notes

Why standard sequence and motif searches miss most of the functional fold space: default tantan repeat masking in DIAMOND masked 86.8% of a structured repeat query in our calibration benchmark, erasing a bona fide 58.8%-identity full-length homolog (E = 1.1×10−166 unmasked vs. zero hits under default settings), while BLAST SEG and DIAMOND tantan disagreed by 20× on masked residues. Worse, screening by 1D sequence motifs or backbone spacing alone routinely selects backbone-only decoys that lack the catalytic or binding side-chain triad in 3D space. Unmasked 3D search paired with explicit 3D side-chain gating solves both blind spots.

Fee $6,000 – $14,000
Turnaround 7–10 business days
Deliverable Natural + ancestral 3D panel

Fee depends on protein family size, whether broken metagenomic hits need full-gene reconstruction (Service 05), and how many ancestral designs are modeled ($6,000 for distant 3D homolog search + diversity-compressed plate panel; $10,000–$14,000 with full ancestral tree reconstruction and 3D refolding). Water and heat stability simulation (Service 01) or repeat-free gene design (Service 02) can be bundled on the final shortlist.

What this does not tell you

Percent sequence-identity tables and 3D structure comparisons against reference patents are technical bioinformatics deliverables, not legal freedom-to-operate opinions or legal advice — broad functional claims, process claims, and legal scope questions require patent counsel. While ancestral proteins often tolerate higher temperatures than modern enzymes and our 3D checks filter out misfolded candidates, distant homologs and reconstructed ancestors can differ in substrate preference, rate, or pH optimum and must be tested in the lab.

Scope & order this service →
SERVICE 09 • TRACK 2

Cryptic bioactive peptide, knottin & miniprotein multimer discovery

Standard gene callers skip open reading frames under 100 residues and leave mature bioactive peptides hidden inside larger precursor proteins. If your delivery format caps chain length — ≤85 residues for solid-phase peptide synthesis (SPPS) or ≤110 residues for a single-AAV vector — those skipped regions hold most of the candidates. We mine genomes, transcriptomes, and metagenomes for cleaved prohormone and antimicrobial peptides, graft binding loops onto 6-cysteine knottins, and find short chains that assemble into split-active-site or metal-cage homo-oligomers, clustered into a non-redundant synthesis plate.

When you should not buy this

If you want a high-affinity peptide binder against a standard soluble target and can run phage display or mRNA display, hire a wet-lab display CRO: physical panning tests 109–1012 peptides directly against your target for measured binding. Buy this service when you want natural or disulfide-knotted compact scaffolds from specific biological niches, cleaved cryptic peptides missed by gene callers, or short chains that build full catalytic sites or metal cages across symmetric subunit interfaces — filtered for chemical SPPS synthesis feasibility before you order a peptide library.

How it works and what you receive
Cryptic cassette excision  Scans precursor proteins across genomes, transcriptomes, or metagenomes for prohormone convertase (KR/RR), double-glycine, and C-terminal G-K-R amidation cleavage sites to excise mature 8–60 AA peptides, scored for DPP-4 aminopeptidase half-life, amphipathic/antimicrobial homology, and sequence distance against published peptide patent databases (DRAMP)
6-Cys knottin loop grafting  Grafts target pharmacophore or binding loops into disulfide-locked 30–50 AA inhibitor cystine knot (ICK) microproteins, checking 3D refolding and canonical (1–4, 2–5, 3–6) cystine-knot connectivity
Short sORF homo-oligomers & plate compression  Finds 25–110 AA chains predicted to stay disordered as monomers that co-fold into symmetric C2–C4 or D2 oligomers forming inter-chain split catalytic triads or axial metal cages (Zn2+, Ln3+), paired with SPPS coupling-stall scoring, NCL ligation junction selection, and synthesis-plate diversity clustering

You send

  • Target mechanism, loop motif, metal ion, or source genomes/metagenomes to mine
  • Hard chain-length ceiling (e.g., ≤50 AA single-shot SPPS, ≤85 AA SPPS/NCL, or ≤110 AA AAV)
  • Reference or patented peptide sequences to track percent identity against

We discover & score

  • Cleaved mature prohormone & antimicrobial cassettes missed by ORF callers
  • 3D-verified 6-Cys knottin loop grafts & coupled-folding short homo-oligomers
  • SPPS coupling difficulty, on-resin aggregation, DPP-4 stability & DRAMP identity

You receive

  • Diversity-compressed, synthesis-ready peptide/miniprotein panel (FASTA + monomer/multimer 3D PDBs)
  • SPPS synthesis & NCL ligation scorecard + cleavage/amidation annotations
  • Sequence-identity tables against DRAMP and client-supplied reference patents

Why short-chain discovery requires dedicated 3D and synthesis filters: standard gene callers discard open reading frames under 100 residues, while unfiltered multimer co-folding on short peptides mostly returns generic beta-sheet amyloid stacks or hydrophobic bundles with no closed active site. Filtering for coupled folding, non-amyloid interface packing, and inter-chain catalytic or metal-binding geometry isolates short homo-oligomers that build real active sites across subunit interfaces while staying compact enough for chemical synthesis or single-AAV packaging.

Fee $10,000 – $18,000
Turnaround 7–10 business days
Deliverable SPPS / AAV candidate panel

Fee depends on whether we are mining client-supplied transcriptomes/metagenomes vs. our indexed public catalogs, and how many multimer co-folding or explicit-water simulation runs (Service 01) are included.

What this does not tell you

Three measured limits apply to short peptides, and we state all three up front. First, monomer structure-prediction confidence (pLDDT) has a strong negative length bias on short peptides (r = −0.832 in our own calibration), so short linear peptides cannot be ordered by raw monomer pLDDT alone. Second, our fast 3-mer antimicrobial homology screen recovers 61.4% of known antimicrobial peptide families (reliable down to roughly 70–90% sequence identity, not deeply divergent peptides below ~25% identity), so receptor agonism, MIC, and hemolysis must be measured in wet-lab bioassays. Third, 6-cysteine knottins that fold cleanly in 3D can still form scrambled disulfide isomers during bench oxidative refolding, so orthogonal Cys protection or refolding buffer optimization is often needed at your peptide CRO.

Scope & order this service →
Track 3 · Verify Clones, Assays & IP

Audit engineered strain genomes, 96-well plate assays & biotech patent filings

Use these three services once constructs are built or filings are drafted: verifying on-target integration, copy number, and uncollapsed tandem repeat arrays from Nanopore or PacBio long reads (03), running high-speed automated droplet-freezing, thermal-shift, or clearance assay analysis on DarkBiome Daybreak (04), or auditing draft patent claims and WIPO ST.26 sequence listings before filing (07).

SERVICE 03 • TRACK 3

Clone validation and integration audit

You have Oxford Nanopore or PacBio HiFi long-read whole-genome sequencing data from engineered fungal, yeast or microbial clones and need to know what actually landed in the genome. We align reads and assemblies against a combined host-genome-and-construct reference (minimap2, Sniffles2, SVIM, mosdepth), call structural variants and copy-number variation (CNV), check raw long reads for collapsed tandem multi-copy gene arrays, and report per clone what integrated on-target, what inserted ectopically, how many copies went in, and whether homologous recombination excised repeat segments.

What you send and what you receive

You send

  • Long reads per clone (Nanopore or PacBio HiFi), or de novo assemblies
  • Host reference genome as FASTA, with GFF3 annotation
  • Your construct sequence as FASTA
  • The locus you targeted, and a neighboring gene to use as control

We report

  • Integration sites and the genes they disrupt
  • Structural variant calls at the locus
  • Construct fragment coverage, segment by segment
  • Partial-deletion, loop-out & tandem repeat-array collapse detection

You receive

  • A per-sample PDF summary and self-contained HTML report
  • Spreadsheet tables of every call
  • The intermediate alignment and variant files

Why raw-read repeat-array checks matter: standard long-read and hybrid genome assemblers routinely merge tandem copies of a multi-copy cassette into a single consensus copy whenever repeat units outrun read overlap — raising assembly N50 and leaving mean contig coverage flat, so standard assembly QC reports a clean single-copy clone when multiple tandem copies (or partial tandem deletions) are actually sitting in the raw reads. Counting unique junction and paralog-specific k-mers directly in the raw reads catches collapsed arrays that assembly-level alignment misses.

Fee $2,500 (up to 4 clones)
Additional clones +$500 per clone
Turnaround 5 business days

Base package covers host reference + construct setup and full integration, structural-variant, and raw-read tandem-collapse auditing for 1 to 4 clones on the same host reference ($2,500 total; +$500 per additional clone in the same batch).

What this does not tell you

We analyze sequencing data you already have. We do not run a sequencer, and we do not culture, transform, or handle your strains. This is a long-read pipeline: short-read data alone is not supported, because the junction-spanning reads the analysis depends on do not exist in it. One target locus and one control locus per run. It was built and proven against a filamentous fungal host, so the first project on a new host or locus is quoted as a short validation run rather than sold as routine. We confirm compatibility with your reference, annotation and read type in writing before quoting, not after.

Scope & order this service →
SERVICE 04 • TRACK 3

Droplet freezing & plate assay analysis

Record a 96-well droplet-freezing plate as it cools — using any radiometric IR camera, a thermal camera calibrated against a thermocouple or 0 °C melting plateau, or an overhead visible-light camera paired with a thermocouple log. We run your recording in-house on DarkBiome Daybreak to return per-well freezing temperatures (with neighbor-well frost waves censored), Kaplan–Meier survival curves, Vali (1971) ice-nucleation spectra K(T), and Koop (2000) solute water-activity (Δaw) corrections. Every report also includes a 95% temperature uncertainty budget, spherical-cap critical contact areas, and group comparisons ready for a manuscript or go/no-go decision.

Formats, measurements and what you receive
Accepts (thermal or optical)  Radiometric thermal video from any scientific IR camera (FLIR .seq/.ffv/.csq, InfraTec, Optris, Telops, Seek, or radiometric MP4/AVI/TIFF streams) • non-radiometric thermal/IR video calibrated via the 0 °C water-melting plateau and/or external thermocouple/RTD/DAQ temperature logs • post-assay visible-light video (.mp4, .avi, .mov, .mkv) or time-lapse image stacks paired with a thermocouple/RTD CSV log or stage cooling ramp • optional validated 96-well DSF thermal-shift (Tm) and turbidimetric polymer/substrate clearance kinetics
Automated QC & calibration  Perspective homography well-grid registration with drift compensation • two-point / melting-plateau & thermocouple temperature reconciliation • automatic detection and right-censoring of sympathetic neighbor-to-neighbor frost waves, premature pre-ramp freezing, and droplet evaporation
Returns  Frozen fraction f(T) • T50/T10/T90 with 95% Wilson intervals & GUM k=2 uncertainty budgets • Vali K(T) and dK/dT • Koop Δaw solute-corrected intrinsic ΔThet • BET site density ns(T) • spherical-cap CNT critical contact-disc area π(r* sin θ)2 • log-rank / Mann–Whitney / Benjamini–Hochberg group comparisons

You send

  • Your 96-well cooling recording (any radiometric thermal video, calibrated IR video, or visible-light video / image stack)
  • Thermocouple, thermistor, or cooling-stage CSV temperature log (for optical or non-radiometric thermal recordings, or 2-point thermal calibration)
  • Plate layout CSV: well IDs, sample names, concentrations, droplet volume, and buffer/solute osmolarity

We measure

  • Per-well freezing onset temperature with sympathetic frost-wave censoring & 95% uncertainty budget
  • Kaplan–Meier survival curves, T50, Vali K(T) spectra & dilution-series stitching
  • Koop solute water-activity (Δaw) correction & spherical-cap CNT critical footprint sizing
  • Replicate CV, multi-group statistical tests, dose–response & cooling-rate dependence

You receive

  • Per-well CSV, plus interactive HTML, multi-sheet Excel, and PDF audit reports
  • Publication-ready vector figures (f(T), Vali K(T), plate heatmap, group boxplots)
  • Annotated video overlay showing real-time well detection and freezing calls
  • CF-1.8 NetCDF dataset + SHA-256 provenance manifest for downstream analysis

DarkBiome Daybreak reproduces published parameterizations (Hiranuma et al. 2015 for Illite NX and Budke & Koop 2015 for Snomax) and tracks motion and thermal drift across multi-gigabyte recordings. It also computes the flat spherical-cap contact-disc area π(r* sin θ)2 in Å2, correcting the ~34× full-sphere overestimate common in the literature to test whether a measured T50 fits a monomer or requires oligomerized assemblies (matched-backbone Thr→Val hydration-shell simulation is available as a Service 01 add-on). You can check our arithmetic against published reference curves before hiring us.

Fee $1,500 per recording
Dilution series $4,500 (up to 5 recordings)
Turnaround 3 business days

Single 96-well recording (radiometric thermal, calibrated IR, or visible-light video + thermocouple log): $1,500. Multi-plate dilution series (up to 5 recordings stitched into a continuous Vali K(T) spectrum): $4,500. Optional validated 96-well DSF thermal-shift (Tm) or turbidimetric clearance plate analysis: $750 per plate (or bundled). We do not run the wet-lab assay, supply the camera, or handle physical samples.

What this does not tell you

Daybreak accepts radiometric thermal files directly (FLIR, InfraTec, Optris, Telops, etc.) and calibrates non-radiometric thermal or visible-light recordings whenever you include a thermocouple/RTD temperature log, stage ramp profile, or recorded 0 °C water-melting plateau. Only uncalibrated recordings that have neither embedded radiometric metadata nor a companion temperature log/ramp/melting plateau are declined rather than guessing temperatures. 96-well layouts are validated; 384-well is not, and we will not run it. This is a research analysis service, not an electronic-signature system, and is not validated for 21 CFR Part 11 or GLP.

Scope & order this service →
SERVICE 07 • TRACK 3

Patent claim & sequence listing technical audit

Patent attorneys check legal claim strategy, while the biological sequences, residue numbers, 3D inter-atomic distances, and WIPO ST.26 XML tables inside a biotech filing frequently go unverified against the underlying math and coordinates until prosecution or diligence. Before you file a provisional, PCT, or non-provisional application — or wire an upfront fee on an in-licensed sequence estate — we run a deterministic computational audit across your draft claims, specification, 3D models, and ST.26 XML to catch mathematically empty percent-identity claims, silent public-database sequence and 3D structural collisions, signal-peptide residue-numbering offsets, and DNA translation bugs while they can still be fixed.

When you should not buy this

We are not a law firm, patent agents, or patent attorneys, and nothing in this report is legal advice, claim-drafting counsel, or a freedom-to-operate opinion. If you do not yet have patent counsel, hire counsel first. And if your filing claims a single wild-type protein with no percent-identity genus, no recited active-site residue indices or 3D distances, and an already-validated ST.26 XML file, there is little here to audit. Buy this when your draft specification recites sequence-identity thresholds, conserved motifs, 3D structural coordinates, or synthetic DNA constructs and you want every number verified for your patent counsel before filing.

What we check and what you receive
Claim math check  Triangle-inequality satisfiability proof checking whether simultaneous limitations (e.g., ≥X% identity to SEQ ID NO: A plus required motif M or <Y% identity to reference B) define a mathematically empty set of zero possible proteins
Public sequence & 3D halo sweep  Unmasked sequence alignment and 3D structural similarity sweep of every claimed sequence and ≥X% percent-identity genus against UniProt, UniParc, NCBI nr, PDB, and public metagenomes to catch silent overlap where an identity halo or structural claim reaches into an existing public record
Residue & 3D check  Verifies every recited residue number (mature vs. signal-peptide-included indexing) and 3D catalytic-triad, salt-bridge, or subunit-interface distance against the actual PDB coordinates
DNA & ST.26 QC  Checks every DNA SEQ ID NO for exact amino-acid translation match, unintended internal stops, unrepaired frameshifts, bacterial TTG/GTG fMet start codons mis-translated as Leu/Val, and WIPO ST.26 XML & USPTO PDF compliance

You send

  • Draft patent claims and specification (Word, Markdown, or PDF) under mutual NDA
  • WIPO ST.26 XML sequence listing or raw protein/DNA FASTA table
  • Any 3D structure files (PDB/mmCIF) cited in the specification

We verify

  • Non-empty claim satisfiability & exact sequence counts per claim
  • Public database “identity bridge” and 3D structural collisions inside your claimed genus
  • Signal-peptide offset shifts & 3D coordinate distance accuracy
  • DNA-to-protein translation parity & ST.26 XML schema rules

You receive

  • Line-by-line technical discrepancy table ready to hand to patent counsel
  • Public sequence & 3D structural collision table with safe percent-identity ceilings
  • Verified residue-index map, 3D distance table & clean ST.26 QC log

Why we built this suite: running deterministic verification checks across more than 1,300 protein and DNA constructs exposed bugs that human proofreading and WIPO ST.26 validators never flag — including mature-versus-precursor residue shifts after signal-peptide cleavage, bacterial TTG/GTG start codons translated as leucine or valine instead of formyl-methionine, and percent-identity genus claims whose outer boundary quietly overlapped unannotated UniParc contigs. Every check is deterministic and reproducible.

Fee $4,500 – $9,500
Turnaround 2–3 business days
Scope 1 patent filing / ST.26 XML

Fee scales within the range by the number of SEQ ID NOs ($4,500 for a single filing up to 50 sequences; $6,500–$9,500 for larger 51–250+ sequence listings or multi-docket licensing diligence estates).

What this does not tell you

This is a technical sequence, coordinate, and mathematical verification service, not legal representation, patent prosecution, or a freedom-to-operate opinion, and hiring DarkBiome creates no attorney-client privilege unless engaged through your counsel under their direction. Our public prior-art sweep covers open sequence and patent-sequence divisions (UniProt, UniParc, NCBI, PDB, and public metagenomes, plus any reference set your counsel supplies), not closed subscription patent registries such as CAS Registry or GenomeQuest; exact database versions and dates are recorded in the report.

Scope & order this service →
Diagnostic Engineering Notes

Why standard protein, gene, and assay pipelines fail

Six concrete notes documenting the exact mathematical, 3D structural, and assembly failure modes behind our nine services — from vendor synthesis portals and eukaryotic ER quality control to 96-well classical nucleation geometry and WIPO ST.26 sequence claims. Read all six technical notes →

Note 01 • Service 02 ($4,000)

Why gene synthesis portals reject repeat proteins — and why random codons stall translation

Why pooled-oligo synthesis refuses repetitive blocks (≥20–200 bp), why vendor tools create identical repeats, and how to diversify codons without slow-tRNA stalls or host loop-out.

Read Note 01 →
Note 02 • Services 01 & 08 ($5k–$15k)

3D structural disulfides vs. unpaired ER cysteines, glycan traps & inert-brick oligomers

Separating load-bearing 3D disulfide collars from free ER thiols, catching catalytic-cleft N/O-glycosylation blocks, and filtering high-ipTM multimers that bury their active site.

Read Note 02 →
Note 03 • Service 04 ($1,500)

The 34× full-sphere CNT area overestimate & sympathetic frost waves in 96-well freezing assays

Why using 4π(r*)^2 instead of the flat basal disc π(r* sin θ)^2 inflates required nucleator area 34.2× at θ = 20°, plus frost-halo censoring and Koop solute shifts.

Read Note 03 →
Note 04 • Services 05, 06 & 09 ($3.5k–$18k)

How default SEG and tantan masking erases real homologs & hides cryptic scaffolds

Why compositional masking zeros out periodic homologs above 50% identity, how contig breaks and microbial contaminants corrupt public entries, and why 100-AA ORF cutoffs miss cleaved peptides.

Read Note 04 →
Note 05 • Service 03 ($2,500)

Why long-read assemblers silently collapse multi-copy tandem expression cassettes

How repeat-graph assemblers fold 24 kb multi-copy Pichia or fungal expression arrays into a single consensus copy — and how raw Nanopore/PacBio junction mapping recovers true copy number.

Read Note 05 →
Note 06 • Service 07 ($4.5k–$9.5k)

Empty-set percent-identity claims, signal-peptide offsets & TTG/GTG start-codon bugs

How simultaneous %ID and motif limitations yield zero mathematically valid sequences, why precursor numbering shifts every claimed residue, and how bacterial start codons mistranslate in ST.26 XML.

Read Note 06 →

Looking for a molecule rather than a service?

These nine fixed-fee services work on your sequences and recordings. Separately, we run our own discovery programs and license them for evaluation, across plastic depolymerases, ice-nucleating proteins, critical-mineral recovery and more. Those are a different conversation, with a different agreement.

Browse the portfolio → Licensing terms →

How a project runs

Mutual NDA first — and early sequences are protected either way.

We will sign your mutual NDA or send ours before you share sequences. Nothing needs to leave your side until it is in place — and if you ever paste or attach a sequence early by mistake, we treat it as strictly confidential from first contact: never added to our corpus, never shared, and deleted on request.

Scope and fixed fee in writing before work starts.

One page: what you send, what you get, when, and for how much. Fixed fee, not hourly. No change in price without a change in scope you agree to.

Half on signature, the balance net 15 on acceptance.

You are not paying in advance for work you have not seen (and for Synthesis Rescue, nothing is billed until your vendor's order-intake screen accepts the redesigned construct). Liability is capped at the fees paid under the project. Your counsel will recognize the term, and we will look at redlines.

Your intellectual property stays yours.

The deliverable is assigned to you on payment. We take no license to your sequences and claim no interest in your results, and your sequences are not added to our discovery corpus, scored against our own programs, or used to train anything. The one thing we keep is what we walked in with: our own tools, methods and anything we had already built. You get the output and full freedom to use it; you are not buying the machinery, and we will say which parts are ours in the scope document rather than leaving it to be argued later.

Local by default. Cloud only if you ask for it, in writing.

By default your sequences stay on our own 48 GB Apple M5 Pro workstation: no shared cloud queues, no public prediction APIs, no third-party AI services, and bundled offline codon and structure databases. Folding, host checks, water simulation, repeat-gene redesign, clone integration analysis, freezing video processing, gene reconstruction, 3D homolog search, patent sequence/claim checks, scaffold optimization, and peptide cassette mining all run locally. Longer simulations or large multi-chain folding jobs finish faster if we add rented GPU capacity, but that means your sequences reach that provider, so we will not do it silently. We quote both options, you pick, and the choice goes in the scope document. If you say local only, the longer runs simply take longer.

Deletion on request.

Tell us when you want your inputs and intermediates destroyed and we will confirm in writing when it is done. Otherwise we hold them for ninety days so you can ask follow-up questions, then delete.

One project at a time, plus academic and expedited options.

A single operator on a single workstation, so your project has the machine to itself. If a slot is not available we will tell you when the next one is rather than queue you. For bounded computational deliverables (Services 02, 04, and 07), an optional 48-hour expedited slot (+35%, auto-waived if missed) is available when the workstation is open; for GPU-intensive screens (Services 01, 06, 08, and 09), optional dedicated cloud GPU acceleration is quoted at pass-through cost with your written opt-in. University and non-profit research labs ordering from a .edu or institutional address receive a 15% academic rate.

DarkBiome LLC (DarkBiome / Dark Biome) provides computational design and analysis services. We are not a contract research organization, we do not operate a wet laboratory, and we are not a law firm or patent-agent practice. Nothing offered here is legal advice, a freedom-to-operate opinion, or a clinical, diagnostic, or regulatory service, and no output should be used as the sole basis for a legal or safety decision. Simulation results are predictions about a model of a molecule, not measurements of the molecule itself.

Common questions

How much does computational protein stability screening cost?

A fixed fee of $5,000 to $10,000 for a screen of 20 to 50 sequences, depending on simulation tier (0.5 ns Triage, 10 ns Standard, or 50 ns+ triplicate Extended), sequence length, and whether multimer interfaces, co-folded ligands, or coordinated metal sites are included. Triage turnaround is 3 to 5 business days.

Can you screen AlphaFold, Boltz-2, OpenFold 3, ESMFold, RFdiffusion, or ProteinMPNN designs before synthesis?

Yes. Send either raw FASTA sequences or 3D structure files across any enzyme, binder, repeat protein, biomaterial, peptide, or biologic — including giant proteins up to 2,000 residues and multi-chain homo-oligomers. Before running water and heat simulations, we check start/end boundaries, 3D structural disulfides versus unpaired ER cysteines, catalytic-cleft glycosylation traps, chemical PTM/protease/MHC-II liabilities, peptide SPPS synthesis stalls, and whether multi-chain designs actually close an active site rather than packing as an inert brick.

What do I do when a gene synthesis vendor rejects my repetitive protein sequence?

Free vendor codon tools reuse the same preferred codon on every repeat — causing vendor rejection and DNA loop-out in the host — while random codon scrambling inserts slow codons that stall translation. Under Service 02 ($4,000 per construct), we vary synonymous codons cleanly across repeats without using slow codons, clear start-site hairpins, hidden splice sites, and cloning-enzyme sites across bacterial, yeast, mammalian, plant, filamentous, or cell-free hosts, and create distinct DNA copies for multi-copy strains. No deposit is due on signature, and the fee is payable only once your vendor's order-intake screen accepts the construct.

What thermal or optical camera formats does DarkBiome Daybreak accept for freezing assay analysis?

Under Service 04 ($1,500 per 96-well recording, or $4,500 for up to 5 recordings in a dilution series; 3 business days), DarkBiome Daybreak accepts radiometric thermal video from any scientific IR camera (FLIR .seq/.ffv/.csq, InfraTec, Optris, Telops, Seek, or radiometric TIFF/MP4/AVI streams), non-radiometric thermal/IR video calibrated via the 0 °C water-melting plateau and/or external thermocouple/RTD logs, and standard post-assay visible-light video or time-lapse image stacks paired with a thermocouple/DAQ CSV temperature log or cooling-stage ramp.

Why do promising database hits so often fail to express or trace back to the wrong organism?

Public genome and metagenome databases are full of incomplete assemblies where sequences were cut off mid-gene, start codons or short exons were miscalled, required co-operon folding chaperones were left behind in the genomic locus, default low-complexity filters masked repetitive homologs, or a hitchhiking microbe was mislabeled as a plant, marine, or eukaryotic gene. Ordering the raw database entry means synthesizing half a protein or sending a bacterial protein into the wrong secretion pathway. Under Service 05 ($3,500 to $7,500, 5 business days), we rebuild the full intact gene from raw sequencing reads, check the flanking operon, verify the complete 3D fold, and confirm the true source organism before you order DNA.

How do you find low-identity alternatives to a patented protein?

Standard sequence search misses relatives below roughly 25% sequence identity — and default low-complexity masking hides repetitive homologs even above 50% identity — leaving R&D teams stuck in the same crowded sequence space as existing patents. Because 3D active-site shape stays conserved long after amino acid sequence diverges, Service 06 ($6,000 to $14,000, 7 to 10 business days) searches by 3D structure across public genomes and metagenomes, reconstructs ancestral versions from the family tree, and compresses the hits into a maximally diverse 24-, 48-, or 96-well synthesis plate with percent identity tables for your team and patent counsel (computational sequence comparisons, not legal opinions).

What does a technical patent claim and ST.26 sequence audit catch before filing?

Patent attorneys review legal claim strategy, while sequence listings and structural recitations often go unverified against the underlying math and coordinates. Under Service 07 ($4,500 to $9,500, 2 to 3 business days), we check draft claims and ST.26 XML files for empty-set math contradictions (where percent-identity and motif rules allow zero possible sequences), silent sequence and 3D structural overlap with public UniParc/GenBank/PDB records, signal-peptide residue-numbering offsets, mismatched 3D active-site distances, DNA translation or bacterial TTG/GTG start-codon errors, and ST.26 formatting defects. It is a computational verification report for your team and patent counsel, never legal advice.

Can you remove unpaired cysteines, deamidation sites, or aggregation patches without breaking our enzyme’s active site?

Under Service 08 ($8,000 to $15,000, 5 to 7 business days), we lock your catalytic triad, cofactor pocket, and subunit-interface contacts in 3D space while optimizing exposed scaffold positions to scrub 3D-unpaired ER cysteines, catalytic-cleft glycosylation traps, NG/DG chemical degradation sites, protease clips, MHC-II immunogenic cores, and amyloid aggregation patches. Every delivered variant (10 to 25 sequences) is independently re-folded in 3D and checked in OpenMM explicit water at 300 K and 330 K before delivery; catalytic rate and titer must still be confirmed at the bench.

How do you discover short peptides, knottins, or sORF homo-oligomers small enough for chemical SPPS or single-AAV vectors?

Standard gene callers skip open reading frames under 100 residues and leave cleaved interior peptides buried inside larger prohormones. Under Service 09 ($10,000 to $18,000, 7 to 10 business days), we mine cleaved prohormone and antimicrobial peptide cassettes, graft pharmacophore loops onto 6-cysteine knottin scaffolds, and identify short 25 to 110 residue chains that self-assemble into symmetric homo-oligomers with inter-chain active sites or metal cages, scored for chemical SPPS synthesis feasibility and clustered into a non-redundant synthesis plate.

Is the shortest tier informative?

For triage, yes. A short explicit-solvent run with a thermal ramp identifies candidates whose predicted fold does not survive solvent and temperature, which in a typical screen is a substantial fraction. It is a filter, not a stability measurement, which is why it is the entry tier rather than the only one. Where the question is whether a specific candidate will hold, Standard or Extended applies.

What if no candidate passes?

You receive that result with the supporting evidence, and the synthesis spend is not incurred. We also identify where in the construct the instability appears, which is generally more actionable than the verdict on its own.

Is there a conflict with your own programs?

Worth stating plainly: we run discovery programs of our own and license the results. Client work is a separate project with separate storage. If your target overlaps one of our programs we will disclose it before you send anything, and you decide whether to proceed.

Do you take single-sequence work, and how does publication attribution work?

Yes, we take single-sequence work — small projects are priced on the same basis and a single-sequence question is often answerable within a day. On publications: you have no obligation to publish (commercial teams keep results private all the time), and we never require co-authorship on paid fee-for-service work. When you do publish results, sequences, or figures generated by a DarkBiome service in a paper or preprint, standard scientific attribution in Materials & Methods or Acknowledgments (citing DarkBiome LLC, Alameda, CA, just as you cite a synthesis vendor or instrument software) applies across all projects — and every deliverable includes a paste-ready Methods section so attribution takes zero extra writing. University and non-profit labs ordering from a .edu or institutional address also receive a 15% academic rate.

Configure & order a service

Select your service, package tier, batch size, and NDA/compute preferences below to see an immediate fixed-fee and turnaround estimate, then submit your scope request. We reply within two business days with a written 1-page scope, fixed fee, and start date.

STEP 01 • CONFIGURE SCOPE

Choose your service, package tier, and approximate batch size. No sequences needed in this form.

STEP 02 • NDA & WRITTEN QUOTE

We reply within two business days with a 1-page fixed-fee scope, next workstation slot, and mutual NDA.

STEP 03 • TRANSFER & DELIVERY

Once you approve the scope, you transfer inputs and work begins. Nothing is billed before you sign off.

1. Service & Package Which computational service and tier do you want to scope?
2. NDA, Compute & Schedule Confidentiality, hardware residency & turnaround options
3. Contact & Problem Summary Where should we send your written scope & quote?
Early-sequence confidentiality pledge: You do not need to send sequences before a mutual NDA is signed. However, if you accidentally paste or attach sequences, coordinates, or assay files early, DarkBiome treats them as strictly confidential from first contact — never added to our discovery corpus, never shared with third parties, and deleted immediately on request.

We use these details to respond to your request and, once it is sent, link it to the record of your visits to this site. See our privacy notice (opens in a new tab).