How to Design an HDR Donor: Arms & PAM Blocking
13 min read · Updated September 11, 2026
An HDR donor is easy to concatenate and hard to choose. The sequence itself is left homology arm, intended edit and right homology arm. The decisions that control whether it is a sensible experiment come earlier: which guide cuts close enough, whether the donor should be an ssODN or a longer template, how long and asymmetric the arms should be, which strand to order, and whether the repaired allele must also carry a blocking mutation.
SeqBench can construct and audit the sequence mechanics, but it does not predict HDR efficiency or select the best ssODN polarity for a cell type. This guide separates the deterministic construction from the context-dependent biology, with a worked 153 nt donor and the checks to make before ordering.
Start with the donor's exact sequence anatomy
For a local replacement, the donor is [left homology arm][replacement][right homology arm]. A substitution replaces one or more reference bases with new bases. An insertion removes nothing and places the new sequence between two coordinates. A deletion has an empty replacement. The arms match the unedited locus on either side and give the repair machinery the homology that places the edit.
That construction is deterministic once the edit interval and arm boundaries are fixed. It does not decide whether the edit preserves a reading frame, whether an epitope tag has the right linker, whether the cell will use HDR, whether the donor will be toxic or whether the chosen guide is active and specific. Those are separate claims and should remain separate in the result.
Use the complete locus sequence actually present in your cells, not only a reference transcript around the coding change. A polymorphism or cell-line variant inside an arm reduces literal homology, and an isoform coordinate can place the right edit into the wrong exon. Verify the reference allele and enough flanking sequence before generating an order string.
Choose the cut before drawing the homology arms
The distance between the nuclease break and the desired change is one of the strongest design variables. Paquet et al. observed a marked decline in precise editing as the programmed mutation moved away from the Cas9 cut in human cells (Nature 2016;533:125-129, doi:10.1038/nature17664). That supports a practical preference for close cuts, not a universal success boundary at one number of bases.
Guide activity still matters. A poorly cutting guide two bases away can lose to a strong guide farther away, and a large systematic ssODN study found useful edits with guides 6-14 bases from the change when guide performance was good (Schubert et al., Sci Rep 2021;11:19482, doi:10.1038/s41598-021-98965-y). Shortlist several nearby guides, screen specificity, and weigh cut distance together with evidence about guide activity.
For a standard 3′-PAM SpCas9 guide, the canonical blunt cut is three bases upstream of the PAM. SeqBench expects the forward-strand coordinates of the protospacer alone: a plus-strand guide cuts after guideEnd − 3, and a minus-strand guide after guideStart + 2. A one-base coordinate error moves the donor's insertion point, both arms and every downstream interpretation together.
Decide whether the donor is an ssODN or a longer template
Small substitutions, short insertions and short deletions are commonly attempted with synthetic single-stranded oligodeoxynucleotides. Larger payloads may need long ssDNA, a PCR-derived double-stranded donor or a donor plasmid. There is no sequence-length switch in the designer that turns one chemistry into another: it returns the bare donor sequence you ask for.
That boundary matters for ordering. An ssODN decision includes strand polarity, purification, synthesis length and any terminal modifications. A dsDNA or plasmid donor includes how the molecule is produced, whether backbone sequence is present and how the homology arms are assembled around the payload. SeqBench does not choose a synthesis platform or add vector architecture.
Do not call any short sequence an ssODN merely because it fits in an oligo order. A 120 nt donor can be ordered in either polarity, but the two orders are reverse-complementary 5′→3′ sequences, not the same string. Purification and end chemistry are independent ordering choices, and SeqBench displays only the forward target orientation.
How long should ssODN homology arms be?
Symmetric arms of roughly 30-60 nt per side are a practical starting range for many small ssODN edits, and 40/40 is a defensible baseline when there is no locus-specific evidence. Paix et al. found that about 35 bases of local homology could support repair in their systems, while the Schubert study used 40 nt symmetric arms as a broad baseline (Proc Natl Acad Sci USA 2017;114:E10745-E10754, doi:10.1073/pnas.1711979114).
Those are starting designs, not universal optima. Cell type, nuclease, edit position, strand, delivery and locus all interact. Longer arms consume synthesis length and can add structure or errors without guaranteeing more repair. Short arms can be inadequate, especially when the available sequence is wrong or one arm is clipped by the submitted locus boundary.
The SeqBench designer takes one armLength and applies it to both sides. If the target sequence does not extend far enough, it clips that arm and reports the actual coordinates and a soft gate warning. It cannot intentionally make asymmetric arms, and a successful returned object does not convert a truncated arm into a complete one.
Symmetric and asymmetric donors are context-dependent
Richardson et al. reported improved editing in tested human-cell contexts with an asymmetric donor carrying 36 bases on the PAM-distal side and 91 on the PAM-proximal side, complementary to the strand released first after Cas9 cleavage (Nat Biotechnol 2016;34:339-344, doi:10.1038/nbt.3481). That result established a mechanistic design strategy; it did not establish 36/91 as a universal recipe.
Later studies found strong locus and system dependence. Lanza et al. saw little overall advantage from arm asymmetry across their mouse conditional-allele experiments, and larger datasets found strand preferences that changed with edit position and guide (BMC Biol 2018;16:69, doi:10.1186/s12915-018-0529-0). A symmetric donor remains a reasonable baseline when the system has no validated asymmetry rule.
SeqBench currently builds symmetric arms only. If evidence supports an asymmetric design, use its reported locus coordinates to assemble the unequal arms explicitly and recheck the resulting sequence. Do not enter the longer value and assume the tool shortened one side according to a Richardson rule; it will make both sides the same requested length.
Which ssODN strand should you order?
There is no context-free answer such as “always order the non-target strand.” Strand labels are also easy to misuse: some authors name the strand paired by the guide RNA, others name the strand whose written sequence appears in the donor, and vendor diagrams can swap “sense” and “antisense” with gene orientation. Define the actual 5′→3′ sequence instead of relying on the label.
Repair can be polarity-sensitive, and the preferred donor can change with whether the edit lies PAM-proximal or PAM-distal, the local sequence and the experimental system. Richardson et al., Kan et al. and Schubert et al. all provide evidence for strand effects, but not one polarity that wins everywhere (Genome Res 2017;27:1099-1111, doi:10.1101/gr.214775.116). Testing both polarities is defensible when the cost and assay permit it.
The HDR Donor Designer emits the donor in the same forward orientation as the target sequence you supplied. It does not choose an ssODN strand. If you decide to order the opposite polarity, reverse-complement the entire finished donor — arms, intended edit and blocking mutation together — after completing every coordinate and translation check on one consistent representation.
Block re-cutting only after checking what the intended edit already changes
A repaired allele that still contains an active PAM and a well-matched protospacer can be cut again, turning a precise repair into an indel substrate. First ask whether the intended edit already disrupts the PAM or sufficiently changes the protospacer. If it does, an extra blocking mutation may add no benefit and may add an unwanted biological change.
When PAM blocking is enabled and a Cas9-family guide is supplied, SeqBench attempts to change one constrained PAM base. It applies the change only if the expected PAM is present, lies outside the edit window and a candidate fits inside the donor span; with a forward-strand CDS frame, synonymous candidates rank first. For an NGG PAM it also warns when a transition leaves NAG or NGA, because SpCas9 can retain activity at near-cognate PAMs (Hsu et al., Nat Biotechnol 2013;31:827-832, doi:10.1038/nbt.2647; Zhang et al., Sci Rep 2014;4:5405, doi:10.1038/srep05405).
A one-base PAM change does not guarantee protection from re-cleavage. Schubert et al. found that the second or third G of NGG often made the strongest single blocking position in their assays and that two blocking changes could be more robust in some contexts. SeqBench does not design seed-region blocks, add a two-mutation strategy or predict residual cleavage.
If the intended edit overlaps the PAM, the tool adds no separate mutation and asks you to verify recognition yourself. It does not simulate the edited PAM before making that decision. If no CDS frame is supplied, it cannot call a change synonymous; if no synonymous candidate exists, a non-silent choice can be returned and must be reviewed as a protein change.
Enter the edit and guide coordinates without changing conventions
Explicit edit coordinates are 1-based and inclusive on the cleaned target sequence. A replacement removes editStart through editEnd and inserts the new bases. An empty replacement is a deletion. For a pure insertion, set editEnd to editStart − 1 so no reference base is removed.
Guide coordinates are also forward-reference coordinates even for a minus-strand guide, and they cover the protospacer without the PAM. The tool recognises SpCas9, SpCas9-NG and SaCas9 cut geometry for guide-derived insertions and PAM placement. It does not check that the interval has the expected protospacer length, so a numerically valid but miscopied guide can still anchor the wrong cut.
Cas12a is deliberately different. Its staggered 5′-PAM cleavage cannot be reduced to the Cas9 blunt-cut rule used here, so automatic cut-derived placement is refused. Supply an explicit edit window for a Cas12a experiment, and design any PAM or seed blocking separately; the current implementation does not do Cas12a blocking.
FASTA headers, whitespace and non-letter separators are removed before coordinates are applied. Confirm the cleaned sequence and avoid multi-record input: concatenating two records creates an artificial junction and makes every coordinate after it look valid while referring to a molecule that does not exist.
Worked example: a 153 nt donor with a PAM block
The built-in example uses a synthetic 399 bp coding sequence, a plus-strand SpCas9 protospacer at positions 180-199, a 33 bp insertion and 60 bp symmetric arms. The guide's TGG PAM is at positions 200-202, and the canonical cut is after position 196, so the insertion is represented as editStart 197 and editEnd 196.
The donor is 60 + 33 + 60 = 153 nt. Its left arm is positions 137-196, the insertion is GTGAGCGGCTGGCGGCTGTTCAAGAAGATTAGC, and the right arm is positions 197-256 after blocking. With frame start 1 supplied, the selected G201C PAM change converts CTG to CTC, preserving leucine in this synthetic frame.
The first returned genotyping pair is CCATAAAAAGCTCCCCTCCA and CTCACGATGATCGTAACCCA, with reported Tms of 60.2 and 60.0 °C. It spans a 138 bp wild-type product; a clean 33 bp insertion makes the expected edited product 171 bp.
Those numbers prove the coordinate arithmetic for the example, not that the donor will edit a cell. The displayed 153 nt donor is in the forward target orientation. The tool did not decide that this is the better ssODN polarity, and the primer pair was searched only on the submitted locus rather than against a whole genome.
Genotyping primers and the PASS badge have narrow meanings
Optional primers are designed on the original unedited target, with a product forced to span the edit. Candidate primers are 18-25 nt, target 57-63 °C around 60 °C, use 40-60% GC and return products from 100 to 1,000 bp. The expected edited product length is the original product minus removed bases plus inserted bases.
That pair can show a sufficiently large size change or provide an amplicon for sequencing. It does not prove both HDR junctions, distinguish a one-base edit by ordinary agarose, rule out partial donor copying, detect random integration or establish genome-wide primer specificity. Design junction-specific or outside-arm assays when those are the claims the experiment needs.
The tool's arm-completeness and primer-found checks are soft. It has no hard gate checks, so any successful run has an overall PASS even when an arm is short, primers were not found or PAM blocking failed. Read every warning and the unchecked list; PASS here means the function returned a mechanically described design, not that the oligo is ready to order.
A pre-order HDR donor checklist
- Verify the target sequence, reference allele and cell-line variants across both proposed arms.
- Compare nearby guides using cut-to-edit distance, activity evidence and the appropriate off-target search.
- Choose donor format, strand polarity and symmetric or asymmetric arms from evidence relevant to the cell type and edit.
- Write the edit in one coordinate convention and independently confirm the cut, removed interval and inserted sequence.
- Check whether the intended edit already disrupts the PAM or protospacer before adding another mutation.
- Translate every coding change, including the PAM block, in the correct strand, frame and genetic code.
- Check the complete ordered donor for length, synthesis constraints, repeats and structure; add modifications only from a protocol you intend to follow.
- Design assays that distinguish precise HDR from wild type, NHEJ, partial incorporation and random integration as required.
- Sequence the edited locus and confirm accuracy-sensitive outcomes independently; no donor-construction tool predicts the HDR percentage.
Frequently asked questions
How long should ssODN homology arms be?
Roughly 30-60 nt per side is a practical starting range for many small edits, and symmetric 40/40 arms are a defensible baseline. It is not a universal optimum: cell type, locus, guide, edit position, strand and delivery matter. SeqBench uses one equal arm length on both sides and warns if the submitted locus clips either arm.
Which ssODN strand should I order?
No one polarity wins in every context. Strand preference can change with edit position, guide and experimental system, and “target,” “non-target,” “sense” and “antisense” are used inconsistently. SeqBench outputs the donor in the target sequence's written orientation and does not choose polarity. If you choose the other strand, reverse-complement the entire completed donor.
Should I always choose the guide closest to the edit?
Prefer a short cut-to-edit distance while retaining a guide that actually cuts well and has acceptable specificity. Published data support a strong distance effect, but a weak guide two bases away can underperform a strong guide farther away. Treat distance as one major variable rather than a hard threshold or the only ranking criterion.
Do I always need a PAM-blocking mutation?
No. First check whether the intended edit already disrupts the PAM or protospacer enough to prevent re-cutting. If the repaired allele still presents a usable target, a PAM or PAM-proximal block can reduce re-cleavage. A single changed base does not guarantee protection, and this tool does not model residual Cas9 activity.
Is the PAM-blocking mutation guaranteed to be silent?
Only when a valid coding frame was supplied and the returned codon change is explicitly synonymous. Without a frame the protein consequence is unknown, and sometimes no synonymous change breaks the PAM. Review the codon, strand and frame yourself; an overall PASS does not certify a silent mutation.
Can the HDR Donor Designer handle Cas12a?
It accepts a Cas12a experiment only when you provide an explicit edit interval. It refuses to infer one cut from Cas12a's staggered cleavage and does not design a Cas12a PAM-blocking mutation. Work out the intended repair geometry and blocking strategy separately rather than borrowing the SpCas9 rule.
Does the donor designer predict HDR efficiency?
No. It constructs arms and edits, checks coordinate and length mechanics, proposes one PAM change and can design local genotyping primers. HDR efficiency depends on the cell, locus, guide activity, donor format, polarity, delivery, cell cycle and protocol. Measure the outcome with an appropriate genotyping assay.
What does PASS mean for an HDR donor?
Very little by itself. The current tool has no hard gate checks, so every successfully returned design has an overall PASS. Arm completeness and primer availability are soft checks, while strand choice, asymmetry, PAM-block success, biology, synthesis feasibility, off-targets and HDR efficiency are not established. Read the individual checks and warnings.
Can I design substitutions, insertions and deletions?
Yes with the explicit edit interval. Replace a 1-based inclusive interval for a substitution or replacement, provide an empty replacement for a deletion, or set editEnd to editStart minus one for a pure insertion. Guide coordinates can still be supplied so a Cas9-family PAM block is considered.
Related references
Related tools
Build a knock-in donor with homology arms around a Cas9 cut site, fold in a PAM-blocking mutation, and get primers that genotype the edit.
Scan a sequence for SpCas9, SaCas9 or Cas12a guide candidates with PAMs and scoring.
Screen a guide's protospacer against curated lab reference genomes — SpCas9 hits ranked by CFD, not by mismatch count.
Enter a template and two primers to predict the PCR product, its size and position.
Put the intended allele itself in the basis: measure what fraction of the pool carries your knock-in, what fraction is still wild type, and what fraction is NHEJ byproduct — from one control trace and one edited trace.
Design SpCas9 pegRNAs for any substitution, insertion, deletion or small replacement — spacer, PBS sweep, RTT and 3' extension, PE3 nicking guides, plus twinPE dual-pegRNA design for large edits.