Record a plate the way you ran it, and the assistant reads your lab's own files
Four new record sheets cover the plates and gels most labs run every week: RT-qPCR, a standard curve, a western blot and an agarose gel. Each one is a form you fill as you work rather than a protocol you tick off, and each names the reporting standard its fields come from, so a reader can judge the result and you can see where every field came from. Separately, when you ask the assistant a lab-wide question it reads your lab's own logs and trackers, not only Plasmole's task board, and names any tab it did not read instead of counting the whole lab from part of it.
New
- Four record sheets for the plates and gels you run every week. Start a notebook entry from a template and pick RT-qPCR plate run sheet, Standards plate, Western blot sheet or Agarose gel sheet. The qPCR sheets follow MIQE 2.0 and the western's antibody table follows MDAR, so each antibody records its catalogue number and RRID, the two a reviewer asks for and a lab can rarely reconstruct a year later
- The gel sheet quotes two GeneRuler ladders word for word from their vendor user guides, with the catalogue and document numbers beside the band list, so you can check it against the vendor's own PDF without leaving the entry. For any other ladder it asks you to copy the band list from its datasheet rather than estimate it
- A plate map inside a notebook entry. Click Plate map in the entry toolbar, or type /plate, name the plate or start from an earlier one, and paint the 96 wells with a sample, a target and replicate numbers. Print gives an A4 sheet that reads on a black-and-white printer, and Save writes the plate's table and a legend into the entry
- No-template and no-reverse-transcriptase controls are now their own well roles on a plate layout. A well that must come back empty is never averaged into a group it is compared against, and never used for background subtraction. On screen and on a printout each carries a marker and a dotted outline as well as a colour, so the two do not depend on telling two tints apart
- The Cloning Designer is open, as a tab. State a construct as ordered parts, paste a list of inserts to see which you already hold and which you still have to synthesise or get hold of, and save the construct with its primers and its lineage. It designs the homology arms your parts do not have and simulates the assembly; nothing it produces has been run at the bench
- Save every construct from the Cloning Designer's Bulk mode at once, into a project and folder you pick. It runs in the background, so you can leave the tab, and Stop halts it after the current row; each row shows what happened to it
- Leave uncut, in the Enzymes tab: name a span of the backbone or a library insert to protect, and the enzymes that open the backbone once without touching it are listed first
- Enzymes are a tab on the plasmid, with the catalogue inline. Run a digest, preview the checks on the map and pin a result to it; Restriction sites and Virtual digest open there, and Virtual gel keeps its own window for lanes from several plasmids. The separate Enzyme Analysis window is gone
- Only me notebook entries in a lab. Mark an entry Only me and nobody else sees it, not another member, an admin or the PI, and it stays out of the lab's search and the assistant's reach, until you share it with the lab
- Saved alignments appear in the Library and on the plasmid page, so a Sanger read you aligned is somewhere you can find it again
- Back and Forward move through the documents you have opened
- Get started is a tab that walks one real job on the sample plasmid every account already has: see what is annotated, design a clone from it, check it before you order, find the one wrong base in a returned Sanger read, and see the design, the plasmid and the verification linked in your notebook. Bring in your own sequence sits at the top from the first minute, and dismissing the tab holds on every device
- Every new plasmid starts with a version 1 that says how it was made, such as created, duplicated or imported, so its history no longer begins in the middle. A save that changes nothing adds no version, and a cloning record names the version and bases of each input that went into the construct
- A plasmid's history names a codon optimization instead of recording it as a plain sequence edit, with the feature and copy recoded, the host and codon table used, how many codons changed, and the CAI either side
- That version also names the other annotations the recode rewrote, so a primer that may no longer bind is not left silent, and says restriction sites were not checked, because nothing in a recode looks at them
- Dictate into a notebook entry: press the mic in the entry toolbar, speak, and the words go in where the cursor is, even inside a table cell. One undo removes them, and a reminder asks you to check names and numbers
- You can also speak to the assistant: the mic in its message box is on
Improved
- The alignment tools are now three, by the question you have. Verify clones checks reads against the plasmid they should match, Align sequences compares sequences, and Assemble reads builds a sequence when there is no plasmid to compare to. Each assembled contig can be saved to the Library, copied, downloaded as FASTA, sent to BLAST or aligned to a plasmid
- The built-in cloning workflow templates carry the reaction conditions their kit manuals state, each checked against the manufacturer's current manual, and the assistant's Golden Gate know-how gives a separate cycling table for BsaI and BsmBI, because they do not run at the same temperature
- The built-in notebook templates describe the sample plasmid Plasmole actually ships, so the expected colony PCR size, the construct and the primer design in the copy match the sequence in your Library
- A notebook table's columns can be resized by their handles, Delete at the end of a cell no longer swallows the next cell, End and Home stay inside the cell you are in, the formula bar keeps its row so the page no longer jumps when you click a cell, and the dialogs that used to leave the page are in the app
- A workbook's text names its dropdown choices and their colours, so a tracker's stages read with its Status column
- Select part of a sequence to design primers against it, invert a linear selection, see Find results as a list, read a digest's overhang bases, and record the fields you need to order a primer
- When you @-mention a picture, such as a gel photo, the assistant looks at it instead of saying the file could not be read
- When the assistant searches your files for one detail, such as a project's current stage, it sees the line that holds it, not only the report's first lines
- When the assistant counts the rows of a sheet, it shows whole cells when there is room, says which column a date filter matched, and when it totals columns whose numbers sit on different rows it says so instead of pairing them
- Bulk mode counts what you paste as inserts and says how many constructs could be designed from them
- When the assistant asks you to approve a notebook entry, a project or an outcome, the card says if the records it read show results short of confirmation, such as a PCR screen, "likely HDR" or sequencing still pending, so you can check its claim before you approve
- A long Research question, such as a protocol with every volume worked out, gets its whole answer instead of stopping partway or ending with none
- The assistant can read up to four lab records in one step, so a question that needs several project reports has more steps left for the answer
- New accounts get a welcome email, and someone who joins from a lab invitation lands in that lab instead of an empty personal workspace
- When you connect Plasmole to ChatGPT, Claude or another AI app, the Allow screen asks once how you found Plasmole. Answering is optional and Allow never waits for it
- plasmole.com/mcp has one-click Add to ChatGPT and Add to Claude buttons, and a first message to paste
- In a lab with AI turned off, the Drive chooser no longer offers what the assistant would do with a file
- The privacy policy names what the browser stores, not only sign-in cookies
Fixed
- A plate reading imported from a BioTek Gen5 export counted a no-template control as a treated well, so an empty control joined the group a comparison averaged. If you compared groups from a Gen5 export with an NTC on the plate, the comparison is worth running again
- Restoring an earlier version of a sheet brought back the older data but left today's plate layout in place, so old readings sat under new well roles, doses and replicate numbers. A restore now brings the layout back with the data
- The codon adaptation index left Met and Trp in the count with a weight of one, so every CAI read a little high, more so for a Met- or Trp-rich gene. It now leaves single-codon families out, as Sharp and Li 1987 define it, so the numbers shown are slightly lower than before and comparable to Biopython's
- Overhang compatibility for an enzyme with a degenerate site, such as StyI, AflIII or AccI, is reported as undecided for that site instead of a wrong yes or no. StyI with NcoI at CCATGG was called incompatible; it is compatible
- Comparing two circular plasmids whose origins differ now turns one to match and says by how many bases, so pUC19 against a copy with the origin moved reads as identical instead of 63% covered. An insertion or deletion is reported as one, never as a moved origin
- When an In-Fusion assembly cannot be built, the refusal now says In-Fusion instead of Gibson
- A Gateway, CRISPR-HDR or homologous-recombination workflow step now tells the genome, the insert and the vector apart from the DNA itself instead of from the order they were wired in, and refuses when both readings fit. Gibson and Golden Gate steps still take their inputs in wiring order
- In the alignment tools: Align with Query on a BLAST hit of the whole plasmid lists the comparison under that plasmid; the notebook note links to the saved comparison; a result tab is named after its comparison and a saved one reopens its own tab instead of a duplicate; deleting a trace asks first, because traces do not go to Trash; the read verdict says what the read shows and your call on the construct sits apart from it; and the plasmid you came from and any plasmid named in the file come first when you assign a read
- In Verify clones, the plasmid you came from stays named when you return to the tab, FASTQ reads get the same plasmid picker as traces, and the one-click Attach appears only when the plasmid you came from and the file name agree, so a read is never filed under whatever was open last
- On the plasmid map's display panel, Hide All and Show All set every layer as they say instead of flipping some back on, a preset applies the enzyme set it names, and Smart Focus reads the open plasmid instead of hiding everything
- The sample entry's enzyme table showed typed-in cut positions that were wrong. It is now computed from the sequence
- After you reject an approval card, the answer says the action was not done, not that it is still waiting for you
- The answer no longer opens with a sentence the assistant wrote before checking your records, such as "Let me view it". That sentence stays with the step it belongs to
- A long question across your whole lab no longer ends in a blank reply. If the assistant runs out of room, it says which calculations it did not finish, or that it ran out of room, and you can try again
- The codon optimizer reads a CDS in its own reading frame. It refuses a gene span, a frame other than codon_start=1, a non-standard genetic code or an internal stop, and shows the whole reason
- The codon panel says when it could not read the sequence as a single coding region, and all four codon surfaces refuse the same cases for the same stated reason
- A recode that narrows to a GC window no longer reorders the shared codon table for everything else
- Apply optimization on a locked plasmid asks to unlock and then applies the result, instead of discarding it
- A recode that would change the protein is refused and nothing is saved. It used to be written and recorded as an ordinary edit
- An optimization waiting on an unlock survives you opening another tab, and asks again when you come back, naming the run it is holding
- Cancelling the unlock closes the prompt instead of leaving it open
- Typing /login reaches the sign-in page instead of a blank one
- When the assistant reads a long spreadsheet a page at a time, each page starts on a whole row, so a cell from one project's row is no longer credited to the next project
- When the assistant cites a file by a garbled id, the chat can now show the file it meant instead of no source
- Security and access-control improvements
