실제로 주문할 수 있는 RNA 설계
단백질, 전사체, 엑손 경계 또는 유전자좌를 붙여넣으세요. 순위와 QC 플래그가 붙은 후보를 받고 파라미터와 한계도 기록됩니다. 12개 후보 풀은 36 크레딧입니다.
- 4
- 설계 종류
- 96
- 실행당 후보
- < 2 min
- 순위 풀까지 시간
- 11
- 설계 레시피
ΔG from a greedy base-pair model — a ranking signal, not a thermodynamic MFE.
설계 레시피에서 시작
각 레시피는 고정 파라미터 세트: 어떤 엔진이 돌고, 무엇으로 거르고, 무엇을 반환하는지.
mRNA vaccine antigen
“Antigen sequence in, a transfection-ready mRNA construct out”
Takes an antigen amino-acid sequence, reverse-translates it against human codon usage, attaches a low-structure 5′ leader and a stabilising 3′ tail, then ranks the constructs by CAI, GC and fold density.
- Candidates
- 6
- Length
- 2400 nt
- Credits
- 58
Reporter construct
“A clean luciferase or fluorescent reporter with a known expression floor”
For assay development: builds a reporter mRNA with a short leader, a humanised CDS and a poly(A) tail, keeping the sequence free of internal restriction sites that would break subcloning.
- Candidates
- 4
- Length
- 1200 nt
- Credits
- 52
mRNA antibody (heavy + light)
“Two chains, one delivery, matched codon usage”
Designs a heavy-chain and light-chain pair that share a leader and a tail, with matched CAI so the two chains express at comparable rates instead of one drowning the other.
- Candidates
- 4
- Length
- 2800 nt
- Credits
- 52
Enzyme replacement
“A metabolic enzyme with a liver-targeted tail”
For metabolic disease programmes: keeps the catalytic domain byte-identical to the reference protein while swapping the leader and tail for hepatic expression.
- Candidates
- 4
- Length
- 3200 nt
- Credits
- 52
Hepatic gene knockdown
“GalNAc-compatible siRNA pool against a liver transcript”
Scans the target transcript with a 21-mer window, filters on GC and seed composition, and ranks candidates by strand asymmetry and internal seed matches — the pool you would send to a GalNAc conjugate vendor.
- Candidates
- 12
- Length
- 21 nt
- Credits
- 36
Oncology target knockdown
“A knockdown pool that survives a tumour-cell transfection”
Tuned for low-transfection-efficiency settings: prefers candidates with higher predicted RISC loading and tolerates a slightly wider GC band than the hepatic recipe.
- Candidates
- 12
- Length
- 21 nt
- Credits
- 36
서열을 위해
Codon optimisation with a real table
Human and E. coli codon usage are both loaded. CAI is computed per construct, not guessed — and you can see which codons were changed.
A UTR library, not a placeholder
Kozak, HBB, albumin leaders and the two-element AES + mtRNR1 tail, each with its length and structure note attached.
Structure estimate you can rank on
A greedy base-pair model reports ΔG estimate, pair density and longest stem, so you can sort candidates by how much structure they carry.
Seed-level off-target scan
For siRNA, seed matches elsewhere in the supplied transcript are counted and shown. For sgRNA, we say plainly that whole-genome search needs an external tool.
QC panel on every candidate
GC, homopolymer runs, complexity, CpG, poly(T) for U6, uridine content for IVT. Failing candidates are flagged, not silently ranked first.
Up to 96 candidates a run
A screening library in one pass, with a cross-candidate diversity check so you do not pay for twelve near-identical sequences.
Vendor-ready export
FASTA, GenBank and CSV, plus IDT / Twist / GenScript order-sheet formatting so the sequence does not have to be retyped.
Every run on the record
Each generation is a row with its parameters, cost and output count. When a reviewer asks which sequence you ordered in March, it is there.
API tokens and completion webhooks
On Lab and Core, API tokens let your pipeline submit a target and receive ranked candidates as JSON, and a run-completion webhook notifies your LIMS when a run finishes. Not available on Bench or the free tier.
타깃에서 주문서까지
- 01
타깃 입력
단백질 서열, 전사체, 엑손 경계 또는 유전자좌 창. 입력 유형이 엔진을 결정합니다.
- 02
풀 생성
코돈 최적화, UTR 조립, 21-mer 스캔, PAM 탐색, 갭머 타일링. 실행당 최대 96개.
- 03
점수보다 QC를 보세요
GC, 동종중합체, 복잡도, CpG, poly(T), 우리딘, CAI, 폴딩 밀도. 엄격 필터 실패 시 플래그와 순위 하락.
- 04
내보내고 주문
FASTA, GenBank, CSV 또는 공급사 주문서. 실행과 선택 변이체가 기록됩니다.
스프레드시트·챗과 비교
| Spreadsheet | Chat window | rnaprog | |
|---|---|---|---|
| Turnaround for 12 candidates | 2–4 days, one person | 20 minutes of prompting, unverified output | Under 2 minutes, scored and QC-flagged |
| Codon table | Whatever spreadsheet is on the shared drive | Usually invented | Published human and E. coli usage, cited in the UI |
| Quality control | Manual eyeballing of GC | None unless you ask | GC, homopolymer, complexity, CpG, poly(T), uridine |
| Reproducibility | Depends who ran it | No parameter record | Every run stores its parameters, cost and output |
| Honesty about limits | Tacit knowledge | Confident and wrong | Heuristic folding and seed-level off-target are labelled as such |
| Export format | Copy-paste into an order form | Copy-paste, with line breaks | FASTA, GenBank, CSV, vendor order sheet |
rnaprog로 설계하는 사람들
Vaccine antigen teams
You have the protein, you need six mRNA constructs by Friday, and the difference between them should be codon choice rather than protein design.
Antigen → 6 constructs → pick 2 for IVT
RNAi therapeutic groups
A GalNAc conjugate programme needs a 12-mer pool per transcript, screened for seed matches and immune motifs before anyone orders oligos.
Transcript → 12 siRNAs → order 3
CRISPR screening labs
A knockout library needs guides that cut early, avoid poly(T), and sit inside constitutive exons. Doing that by hand does not scale past ten genes.
Gene panel → guides → library order
ASO and splicing programmes
Exon-skipping chemistry needs boundary coverage and a Tm window, and a gapmer that blocks rather than cleaves needs a different score than a knockdown gapmer.
Exon boundary → 12 gapmers → 3 to test
Synthetic biology and strain engineering
Switching host from human to E. coli changes which codons are optimal. One toggle, and the same protein comes back with a different CDS.
Same protein, two hosts, two CDS
CROs and design services
Client sends a target, you send back a ranked pool with a QC table and a pre-synthesis report. The audit log is what makes it billable.
Client target → report → synthesis order
예시 실행
전체 보기 →Rabies glycoprotein, six constructs
Reverse-translated against human usage, six variants differing only in synonymous codons. Two were carried forward to IVT; the rest were dropped for fold density above 18%.
- Variants
- 6
- CAI range
- 0.86 – 0.94
- Credits
- 46
Illustrative run using a public-domain sequence.
Hepatic target, 12-mer pool
Four candidates were dropped for a UGUGU motif and three for a seed match inside the transcript. The top three were ordered as GalNAc conjugates.
- Candidates
- 12
- Dropped
- 7
- Credits
- 48
Illustrative run on a reference transcript.
안 쓰는 좌석이 아니라 연산에 지불
크레딧이 유일한 계량 단위. 플랜은 매월 갱신, 팩은 만료 없음.
Bench
A single lab testing a handful of constructs
≈ ¥199 · 1 seat · 96 candidates/run
3,000 credits / month
≈ 65 mRNA constructs or 1,000 candidates
지금 시작- 3,000 credits every month
- All four design kinds: mRNA, siRNA, sgRNA, ASO
- Full sequence + score export (FASTA, GenBank, CSV)
- QC panel on every candidate
- Heuristic folding and seed-level off-target scan
- Email support, 2 business days
Lab
A group running several targets in parallel
≈ ¥699 · 3 seats · 96 candidates/run
12,000 credits / month
≈ 260 mRNA constructs or 4,000 candidates
지금 시작- 12,000 credits every month
- Everything in Bench
- 3 seats with shared projects
- API tokens and run-completion webhooks
- Batch runs up to 96 candidates
- Pre-synthesis report per selected variant
- Priority support, 1 business day
Core
A platform team or a CRO serving external clients
≈ ¥2799 · 5 seats · 96 candidates/run
60,000 credits / month
≈ 1,300 mRNA constructs or 20,000 candidates
지금 시작- 60,000 credits every month
- Everything in Lab
- 5 seats, shared project library
- Vendor-formatted synthesis exports (IDT, Twist, GenScript)
- Run-level audit log for internal review
- Invoice billing on request
- Priority support, 4 business hours
1,000 credits · $19
A single target, end to end
5,000 credits · $79
A screening round
20,000 credits · $279
A full library pass
팩은 만료 없이 플랜 크레딧 소진 후 사용.
| 단계별 비용 | Unit | Credits |
|---|---|---|
| mRNA construct design Codon optimisation, UTR assembly, fold estimate | per construct | 40 |
| Candidate sequence siRNA / sgRNA / ASO, scored and QC-flagged | per candidate | 3 |
| Folding estimate Greedy base-pair model with structure map | per sequence | 6 |
| Off-target scan Seed-level internal scan, not whole-genome BLAST | per target | 12 |
| Pre-synthesis report PDF with QC table, scores, vendor-ready notes | per variant | 8 |
| Library batch Package price including one off-target scan: 96 candidates at 3 credits (288) plus one 12-credit scan = 300. A pure 96-candidate pool with no scan costs 288. | per 96 candidates | 300 |
예: 12개 siRNA 풀 + 오프타깃 스캔 = 48 크레딧.
먼저 묻는 질문
Does my sequence leave my control?
Sequence input is stored in your project record so you can reopen it, and it is never used to train a model. Delete a project and the sequence, candidates and runs go with it. Full export is available on every plan.
Is the off-target analysis a whole-genome search?
No, and we do not pretend otherwise. For siRNA we count seed matches inside the transcript you supply. For sgRNA we report the spacer and its coordinates but a whole-genome search needs an aligner such as Cas-OFFinder or BLAST. Treat our number as a filter, not a verdict.
Is the folding estimate the minimum free energy?
It is a greedy base-pair model, not a thermodynamic MFE from ViennaRNA or RNAfold. It is useful for ranking candidates against each other and for spotting structure-dense constructs. Every page labels it as heuristic.
Can I order synthesis directly?
You can generate a vendor-formatted order sheet for IDT, Twist or GenScript and record the order in the platform. We do not place the order or take a cut of it — you buy from the vendor under your own account.
Which modifications are supported?
The design layer handles 5-10-5 gapmer architecture with 2′-OMe wings and GalNAc-compatible siRNA lengths. Uridine content is reported in the QC panel, but codon choice is not uridine-aware and CDS is not optimised for m1Ψ. Modification chemistry is a vendor decision; we produce the sequence and the architecture note.
Does a high CAI guarantee expression?
No. CAI correlates with expression in some systems and poorly in others. 5′ structure, uridine content, the leader and the tail usually matter more than the last few percent of CAI. We show all of those so you are not optimising one number.
Can this be used for a regulatory submission?
The output is a design starting point with a documented parameter set and an audit trail. It is not a validated design control system, and we do not claim GxP compliance. Your quality system still owns the decision.
What happens if a run fails?
Credits are charged when a run starts and refunded automatically if the run fails before producing candidates. A partial run charges only for what it produced.
Do you keep my sequences after I cancel?
Projects stay readable for 30 days after cancellation so you can export them, then they are removed. You can delete a project at any time and the removal is immediate.
Which hosts are supported for codon optimisation?
Human (default), CHO, E. coli, yeast and cell-free. Human and E. coli use published codon usage tables; CHO and yeast currently reuse the closest available table and are labelled accordingly in the UI.