Using DeepCpf1

Using DeepCpf1 Model_Overview

Load Example

The Load Example button is able to replicate the example inputs shown in this help page for quick testing of the DeepCpf1 web server. Simply click the Load Example button then press Submit Job to obtain an example results page.Chromatin Accessibility Chromatin_Accessibility

Considering chromatin accessibility? If you select Yes (DeepCpf1), you will need to select a cell line by searching and clicking on the row of the table that shows. If you select No (Seq-DeepCpf1), you will not need to select a cell line.Information on cell lines are available in Cell_Line_Information.xlsx.Accessibility is read off the chosen cell line's DNase-seq peaks by aligning each target to the human GRCh38 index, so what the feature answers is whether that genome places the site inside an open region. Any sequence may be submitted with it — the alignment decides, not where the sequence came from — and a target the human index does not place is scored as inaccessible, with Chr (GRCh38) and Pos (GRCh38) reading - on that row. Where the site's accessibility is not the question, No (Seq-DeepCpf1) scores the sequence on its own.InputInput MethodYou can select Enter FASTA Text (sequence length >= 34) or Upload FASTA File. FASTA_Input

When uploading files, press the [+] button to add files. Uploaded files can be cleared by pressing either of the checkmark buttons.Input FormatsNine formats are accepted, and each resolves to the target sequences this web server scores: FASTA (text or file), Gene Symbol, NMID, Ensembl ID, HGNC ID, ClinVar ID, COSMIC ID, Input Sequence (an unedited/edited pair) and Genomic Position. The Input Format dropdown selects one and shows the field it needs.SpeciesA gene identifier is looked up in Human (GENCODE v50 on GRCh38) or Mouse (GENCODE M39 on GRCm39). The selector appears with the four gene formats, and it is offered wherever the genome-wide off-target search offers Mouse, so a run that screens against the mouse genome can design in mouse as well.A mouse symbol from an older nomenclature resolves to the current one and the job log states the substitution — Olfr1 designs against Or1e16. Whichever of the four gene formats is chosen, Mouse replaces the identifier box with Enter Mouse Gene Name or ID, which takes a mouse symbol or a mouse Ensembl gene ID. ClinVar ID and COSMIC ID carry no species selector: both are human variant databases, and both resolve against the human reference.Target Sites: Primary and Secondary PAMsCas12a reads its PAM on the 5' side of the protospacer, and the site search takes TTTV — TTTA, TTTC or TTTG — by default. Setting Secondary-PAM (TTTT) target sites to Yes widens the search to TTTN, which enumerates the fourth PAM this nuclease uses as well.TTTT is a Cas12a PAM rather than a non-PAM. It carries 3,635 of the 15,000 target sites Seq-DeepCpf1 was trained on, it occupies the same four positions of the model's input as any other PAM, and the trained weights hold its cost: over otherwise matched 34 bp targets the model scores TTTT sites at about 34.5 where the three TTTV PAMs score 49.5 to 54.7, and below the matched TTTV control in 99.9% of pairs. A TTTT site is a site this nuclease cuts, and one it cuts less well.A widened run labels every row in a PAM class column: primary for a TTTV site, secondary for a TTTT one. Both classes are ranked together on predicted activity, which puts most secondary sites low in the table, and the column is what says which class a row belongs to. A run left at TTTV finds primary sites only, so the column is absent rather than repeating one value on every row.The choice reaches the rest of the run. The genome-wide off-target search looks for the PAMs the run enumerates, so a widened run counts the TTTT sites a genome holds against a guide's specificity, where a TTTV run leaves them out of the count. A PAM-blocking mutation in a homology-directed repair (HDR) donor has to defeat every PAM the run accepts, so under TTTN it changes one of the three fixed T's.Homology-Directed Repair (HDR) Donor TemplatesThere is no HDR mode switch: the input format decides. When the input format specifies an edit — an unedited/edited sequence pair, a ClinVar ID or a COSMIC ID — every candidate guide is also reported with what it needs for HDR. The other six formats identify the target without specifying an edit, so no donor is designed from them and the donor columns are absent rather than blank.Cas12a does not cut bluntly. It leaves a staggered break distal to the PAM, so both boundaries are reported — Cut site (1-based in the target sequence) and Cut site (opposite strand) — and a donor's arms are measured from the middle of the wound rather than from a single position.Each candidate carries these columns, in the results table and in the downloadable files alike:

Cut-to-edit distance (nt) — how far the break sits from the base being changed. Candidates are ranked by it (HDR donor rank in the target), because repair efficiency falls away with distance; candidates at the same distance keep the order the model listed them in.

Donor (sense) / Donor (antisense) — the single-stranded donor carrying the edit, in the orientation(s) you asked for, ready to order.

Donor length (nt), Left arm (nt), Right arm (nt), Requested arms (nt) — arms are measured outward from the union of the cut site and the edit, not from either alone, so that a guide cutting some distance away still spans both. The realised arms therefore differ from the requested pair whenever the cut sits outside the edit, and both are reported.

Indel length not a multiple of 3 — a length test on the edit itself. It is deliberately not called frame-disrupting: the donor is built from a sequence window, so whether the position is coding at all is not knowable here.

Blocking mutation — off by default. When enabled, a change that stops the nuclease re-cutting the repaired allele.

Donor note — what shaped this donor: the staggered break the arms are measured across, why a blocking mutation was or was not made, and by how much the arms were extended outward past the requested lengths so that the donor spans the cut as well as the edit. Where a row carries no donor, the note gives the reason in its place — arms that do not fit inside the supplied window, naming the extra sequence or the shorter arm that would make them fit, or a window that could not be tied to exactly one input record.

What you can set: homology arm lengths (100/100 nt by default, 50/50, or a custom pair between 25 and 1000 nt), the donor strand, whether to add a blocking mutation, and the donor template type — a chemically synthesized single-stranded oligodeoxynucleotide (ssODN) by default, with the donor note saying when a template outgrows standard oligo synthesis and should be ordered as a long single-stranded DNA (ssDNA) fragment instead; long ssDNA for those longer templates; or a double-stranded DNA (dsDNA) duplex, which arrives as one Duplex donor (dsDNA) column instead of the strand pair. The guide sequence is deliberately not yours to specify — the guides are this model's own ranked candidates, and a donor is built for each of them.Guide Cloning OligosCloning Vector decides whether the results carry an ordering-ready oligo pair. Two Cas12a vectors are offered — Cas12a / pY108 or pU6-crRNA (BsmBI), the default, and Cas12a / pRDA_052 Broad GPP (BsmBI) — and every candidate then gains Oligo (forward) and Oligo (reverse), in the results table and in the downloadable files alike: the two strands that anneal into the digested vector, each presenting the 5' overhang that vector needs. A Cas12a crRNA takes no initiating G, so the protospacer is carried as it stands, and it is the full 23 nt protospacer the model scores rather than any shorter view of it.Setting the vector to None (No Oligo Output) leaves both columns out rather than blank. The four overhang fields — Forward 5′, Reverse 5′, Forward 3′ and Reverse 3′ — override what the vector supplies; left blank, each takes the vector's own. A guide carrying anything outside A, C, G and T gives an empty pair, because there is nothing orderable to write.Off-Target ScoreThe Off-target score (higher = less specific) column is a mismatch-tolerance score, not a count of sites in a genome. It asks a narrower question than its name suggests: for this guide, how much activity would survive a single mismatch, averaged over every single-mismatch variant of the protospacer. A guide that tolerates mismatches well is a guide that is more likely to cut somewhere it was not aimed at, which is why a higher number is the worse one.It is computed from the Cas12a substitution table (DeWeirdt et al. 2021, scored by the scheme of Doench, Fusi et al. 2016), so it needs no genome and runs in the time the on-target prediction already takes. It cannot tell you where an off-target is, and it does not know whether any of those variants occurs in your genome at all.For that, set Genome-wide off-target search to Human or Mouse. That search adds one Genome-wide specificity (higher = more specific) column, which runs the other way round: there, higher is better. Every site it finds is condensed into that single number; the individual loci are not listed, so no chromosome, coordinate or strand for them reaches the table. The two are different measurements and both may appear. The sites that search finds are scored on the same table, so the two columns differ only in the question they answer -- one guide against every single-mismatch variant of itself, one guide against the sites a genome really holds -- and never in how a mismatch is weighed.Results Result_1

Users are able to click on each row of the table to view and comprehend the detailed results in the Selected Target Details section above. All columns may be sorted and filtered in the table.

Input / Target Sequence (34bp): This is the 34-base pair genomic DNA sequence used as the input for the deep learning model. It includes the nucleotide composition required for the Cpf1 enzyme to recognize and bind to the site.

Guide Sequence (20bp): The first 20 nucleotides of the protospacer that directs the Cpf1 enzyme to the matching target DNA. The Cas12a protospacer is 23 nucleotides long — 4 bp of 5' flank, the 4 bp TTTV PAM, the 23 bp protospacer and 3 bp after it make up the 34 bp window above — so this column is a display convention over it rather than the geometry. Everything derived from the protospacer uses all 23: the scored window, the cloning oligos and the genome-wide off-target search.

GC Contents (%): The percentage of Guanine (G) and Cytosine (C) bases in the 20 nucleotides shown under Guide Sequence (20bp), rather than across the full 23-nucleotide protospacer. This is a standard feature used in evaluating the stability and efficiency of CRISPR targeting.

Strand: Indicates which DNA strand contains the target sequence; (+) represents the forward or sense strand.

Chr (GRCh38): The specific chromosome where the target sequence is located (e.g., chr21). The column is filled from an alignment against the human GRCh38 index, so a site that index does not place — a site resolved from a mouse gene among them — reads - here and in Pos (GRCh38).

Pos (GRCh38): The specific starting coordinate (base pair position) of the target site on the chromosome, given against the GRCh38 human reference assembly, matching the coordinates used everywhere else on this site.

Location: The relative range of the sequence within a specific gene or reference region.

ID: A unique identifier for the target, often corresponding to the gene name (e.g., LINC00478).

In chromatin-accessibility peak (cell line): Refers to DNase-seq narrow peaks, which indicate chromatin accessibility. The paper demonstrates that considering whether a target site is in a "peak" (accessible) area is essential for accurately predicting Cpf1 activity. A "1" means the site overlaps a DNase peak in the selected cell line, a "0" means it does not, and a "-" means chromatin accessibility was not considered in the prediction (Seq-DeepCpf1).

DeepCpf1 Score: The final output of the DeepCpf1 algorithm. It is a regression score that predicts the expected activity or indel frequency (editing efficiency) of that specific guide RNA. A run made without chromatin accessibility carries the same quantity under the heading Seq-deepCpf1 Score, which names the model that produced it.

Users can download the results by clicking the Download Results button. Download_Results

Sample OutputsThe following example interactive outputs are available for reference:FASTA sequenceGene symbolClinVar variantSequence pairGenomic coordinatesContactIf you have any questions, bug reports, or suggestions, please do not hesitate to contact us through the Support page.

DeepCpf1 web server

About DeepCpf1

All web servers

Your browser does not support ES modules.
Please use a modern browser.