Gene Studio 1.1.50.18

Changes in 1.1.50.18

Gene Studio 1.1.50.18

A licence request now says what it is for, and can carry a work address to
back it. This release changes the request window, the site and the editor's own
known-feature scan; the licence desk behind them changed with it.

• What this is for. The licence request asks whether the seat is for a
  student, for academic research or for a company, the same way it asks for a
  name, a lab and a country. It is asked rather than assumed: the price is read
  off the answer, and a window that guesses is a window that quotes a
  professor the company price.
• A work address, if you have one. Student and academic prices are
  confirmed against an address at your institution. Where the e-mail on the
  request already is one, nothing more is needed; where it is not — a lab that
  writes from Gmail, say — a work address can be added, and on genestudio.app
  a link is sent to it. Following the link confirms that you can read mail
  there, and that is all it confirms; nothing else is looked up. The request
  window itself has no network and confirms nothing — it carries the address,
  and the desk reads it.
• On genestudio.app, the account page asks for a country (from the same
  list the application uses) and a work address, and a request made there
  says what it is for. The administrator's page tells a request that is yours
  to answer from one that is waiting on the applicant, and can remove one that
  was never a request.
• Known features follow the sequence. The scan for known features ran
  when a sequence was opened or pasted and nowhere else, so a construct
  started empty and filled in afterwards — typed, or pasted into the sequence
  view — was never offered anything, and one that had been scanned asked to
  be scanned again after every edit. It now runs by itself once an edit
  settles, on sequences up to 50 kb, and a PCR or cloning product opened as a
  new construct is scanned like an imported one. A bar you have closed stays
  closed while you edit.

Nothing changes for a licence already issued: the signed licence, its machine
binding and its end date are untouched, and a request already waiting is
answered as before. A confirmed work address is evidence of where somebody
works, not of what they are — a university address is held by staff and
alumni as well as students — so the student price remains a person's decision.
Where no work address can be given, the request still goes through; say so in
the note and a person reads it.

Validation includes the full regression suite, the request window measured in
its own 500×560 frame with the two new fields reachable and focusable, the
request composer and the desk's reader held to the same field labels, and the
live site driven end to end against a local copy including the daily digest.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.17

Changes in 1.1.50.17

Gene Studio 1.1.50.17

The local known-feature review gets the three things it was missing: a topology
to work with, genes it had been blind to because they were fused, and references
for the parts it kept not finding. The feature library's limits are now counted
in the unit they are stated in.

• A pasted sequence can say it is a plasmid. The paste dialog had no
  topology control and defaulted to linear, so a plasmid whose origin falls
  inside a gene lost that gene — on pUC19 rotated by 2,000 bp, the missing
  candidate was AmpR. There is a Topology menu now, filled from a record's own
  LOCUS line. Where the sequence is still read as linear, the scan reports only
  the difference closing it would make — "1 more match if this is a plasmid: it
  crosses the origin" — with the button that closes it. The topology is not
  guessed, and the press is one Undo.
• A gene fused at either end is the same gene. A CDS whose start or stop
  codon has been replaced by a tag or a linker was invisible to both searches:
  the exact search wants every base including both termini, and the approximate
  search declined it because the bases that changed are the stop codon. The
  coding body is now searched on its own and reported as the fusion it is, with
  the replaced codon and the sequence that replaced it named. Where the stop
  codon was replaced, the reading is followed to the next in-frame stop, which
  is where that protein actually ends; where there is no stop to reach, the 3′
  end stays open rather than being invented. On the deposited pEGFP-C1 and
  pGEX-4T1 records this puts the candidate on each record's own fusion CDS to
  the base.
• A motif you would expect to find anyway is not evidence. A four- or
  six-residue site is contained by chance in any long sequence, so short motif
  matches with nothing saying the stretch is translated are set aside and
  counted rather than listed as findings. The list says what is holding them
  back — "Every match here is a short motif this sequence is long enough to
  contain by chance" — and offers the tick box that shows them.
• One unread base no longer loses a gene. Both searches required ACGT
  throughout, so a single N — which a Sanger read of a construct routinely
  carries — dropped a 720 nt CDS and the import offered nothing. An N now costs
  what a substitution costs and is counted apart from one, because it is not a
  base that was read and found different.
• The obvious button says what it is leaving behind. "Select visible exact"
  silently skipped every candidate needing review, which on an ordinary vector
  is the pUC origin at 99.8% identity — a map built with that button had no
  origin of replication and nothing said why. Accepting the suggestions also no
  longer reports the construct as changed and asks you to scan again about your
  own press.
• Two more verified references, and the catalog is 38 intervals. The
  complete Schistosoma japonicum Sj26 glutathione S-transferase CDS
  (M14654.1) identifies the GST tag the pGEX vectors carry, where before there
  was only a fifteen-residue peptide motif; with the fusion reading above it
  lands on pGEX-4T1's own annotated fusion CDS exactly. The pGL3 upstream
  poly(A) signal is now its full deposited 154 nt interval (U47298.2) rather
  than the shorter pCI-neo signal that is its first 49 bases, and the two are
  recognised as one element at two extents. The luciferase reference now cites
  the current version of its accession rather than a superseded one; its
  sequence, interval and translation are unchanged.
• The feature library has one limit, counted in bytes. Saving and importing
  weighed text two different ways, so a library whose metadata is written in
  Korean walked through the budget — a Hangul character is one unit and three
  bytes — and the app then refused its own backup. Both paths weigh bytes now,
  an export that would not fit back through the import is refused where it is
  made, and the budget is 32 MiB: the smallest figure that admits the 1,000
  entries the entry cap already allows, so the two limits no longer contradict
  each other. The qualifiers that say a straight reading of a gene's bases is
  not the protein it makes — including /exception and /ribosomal_slippage —
  now survive a save, instead of an entry coming back out as a protein
  reference the matcher itself would refuse.
• A CDS read from its second base keeps the peptide its record and its bases
  agree on. A /codon_start=2 CDS beginning on TTG or GTG was translated two
  different ways in one session: the feature dialog called the record's own
  correct protein a disagreement with its own bases, and an unrelated edit
  inside the CDS deleted /translation and left a note saying the peptide had
  never matched. It had. One answer now serves every reader of it.
• Genes over the origin, in two more places. A gene annotated across the
  join came back as a new candidate because its segments were read as covering
  nothing, and a tag inside a gene whose reading frame shifts mid-way went
  unflagged. Two candidates for the same stretch are also no longer both added.
• The browser and phone builds cannot fall behind the desktop. The shipped
  web copies are generated from the desktop page and are checked against it
  before every commit, so a fix landed here reaches genestudio.app rather than
  shipping as the bug it fixed.

The local reference search does not upload sequence, and no candidate is added
to a construct without being ticked. A fusion candidate and a similar-DNA
candidate are review evidence: a match does not establish that a sequence is the
reference's own allele, that a fusion partner is in frame, or that anything is
expressed. Similar-DNA matching remains substitution-only; a 3′ end is read to a
stop codon only for a fused coding body, and only while the reading stays
shorter than the body it was matched on. In-frame insertions and deletions,
partial CDS boundaries and intron inference are still not attempted. The
reference catalog is small, source-backed and deliberately so; it is not a
complete parts database, and absence from it says nothing about a sequence.

Validation includes the full regression suite, the known-feature engine at
398/398 checks, the five-plasmid benchmark at 60/60 key regions across three
coordinate transforms, the three independent plasmid records whose every
predeclared interval is now covered, the representative CDS catalog at 72/72,
the review dialog at three widths, and a revert check on each fix so that
undoing it names the defect by name.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.16

Changes in 1.1.50.16

Gene Studio 1.1.50.16

Imported DNA can now be reviewed against a local feature library, then passed
on to gene prediction and functional annotation when more evidence is needed.

• Known features on import. Opening or pasting DNA offers local reference
  matches without changing the sequence or adding unchecked annotations.
  Both strands and circular origins are searched. Exact DNA, complete-protein
  identity and conservative similar-DNA candidates are distinguished, with
  reading-frame and short-motif conflicts marked for review.
• More complete CDS references. The verified catalog contains 36 source
  intervals, alongside the existing short motifs. Complete references include
  selectable markers, fluorescent proteins, luciferases, TetR, AraC, Cre, Flp,
  SacB and a specifically identified SpCas9 fusion. Source accession versions,
  boundaries and translations are retained; vector variants keep their labels.
  Protein and similar-DNA matches identify the matched reference without
  claiming its DNA allele. Protein evidence uses a verified protein accession
  when available, and lacI CDS matches no longer inherit the lacIq allele claim.
• Inspect before adding. Candidate details show sequence context,
  translation and actual reference differences, including affected codons.
  Search, feature-type filters, bounded pages and keyboard controls help review
  long lists. A selected batch is added in one undoable edit; existing
  annotations are preserved.
• Keep a laboratory feature library. Save a complete contiguous feature as
  an independent reference, edit its name and DNA/protein matching preference,
  and export or merge libraries as JSON. Conflicting imported definitions are
  reported without silently replacing existing choices. The library explains
  its separate backup requirement and warns about very short references.
• Continue into annotation. The candidate review links to the existing
  prediction tools. Prodigal predictions receive stable protein identifiers,
  retain partial boundaries and coding phase, and avoid adding the same CDS
  again, including predictions that only choose a different start codon.
  Circular predictions use flanking context and fold complete calls across the
  origin; unresolved edge fragments are counted and omitted. Existing CDSs can
  receive missing identifiers in one undoable step so
  eggNOG and InterProScan results can be joined back to the correct features.
  Changed documents and stale results are refused.

The local reference search does not upload sequence. Prodigal runs in the
desktop app with the local tool installed; functional services still require an
explicit submission or results-file import. A match or a predicted CDS does not
establish biological activity. Similar-DNA matching remains substitution-only;
general partial-CDS and fusion-boundary inference are not included.
Circular prediction uses at most 10 kb of context per side, so longer
origin-crossing genes may need another annotation method. Large manual local
reference searches can pause the interface; the import offer can open gene
prediction directly without running that search first.

Validation includes the full regression suite, responsive review and pipeline
handoff checks, public reference sequences and an actual Prodigal 2.6.3 run.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.15

Changes in 1.1.50.15

Gene Studio 1.1.50.15

A deletion is picked by pointing at the genes it takes, and its ends are cut the
way the lab cuts them.

• Point at the region, and it lands on the genes. Under the Target boxes in
  Guides there is now a strip of the molecule around the deletion: the genes in
  it, the bases that go, and — outside them, hatched — the two homology arms
  that will be amplified, because where the arms fall is a consequence of the
  region and is the thing worth looking at before four primers are ordered.
  Drag across it and the ends snap outward to whole genes: the first gene's own
  start, the last gene's own stop, so a deletion that stops two bases inside a
  gene stops being something you can do by accident. A gene's own bar takes that
  gene alone, and under the strip each gene is a tick box — ticking extends the
  stretch through it, unticking an end pulls it back, and a gene in the middle
  of an operon says so rather than quietly doing nothing. A stretch nobody
  annotated is still a stretch: drag over bare DNA and the coordinates stand as
  drawn.

• Whole gene, or in frame — both are offered. One lab deletes a gene start
  codon to stop; another keeps both codons and removes what is between them, so
  the remnant is in frame and what the operon transcribes after it is still
  translated. Deletion ends now carries both, and a third that keeps a set
  number of codons at each end. Which codon an end keeps depends on that gene's
  strand — the low end of a forward gene is its start codon and of a reverse
  gene its stop — so the readout names the gene and the codon rather than
  leaving it to be worked out from an arrow.

• What has to be in frame is the join, not the length removed. The two
  retained pieces are whole codons of their own genes, so the fusion reads
  through exactly when what stays is a whole number of codons — and a
  three-gene stretch whose ends keep a codon each is in frame however many
  bases went between them. Measured on a real pair of reading frames, that
  deletion leaves ATG reading straight into TAA; judging the 1,994 bases
  removed instead would have called it two bases out. Where a junction really
  is out of frame — after a nudge, say — the line says so and a button keeps
  the one or two bases that put it right. Either end can be nudged without
  losing the genes the region was picked on, and the two cuts are drawn at base
  resolution underneath, with the ruler, the gene and the translation, so what
  is kept and what goes is read rather than inferred.

• The tube still says ΔgltA. A design was named by looking for a feature
  whose start and end are the span, so keeping a codon at each end named the
  deletion after its coordinates instead — on all four arm primers, which is
  what the tube carries. The genes the region was picked on are carried beside
  the bases that go, and an operon is named by its first and last gene. The same
  rule travels to a genome-wide run, where it is applied to each CDS and the
  guide is searched inside the bases that actually go — a guide that cut outside
  them would cut the edited genome again.

• The Guides panel stopped jumping. Every press in that view is put back at
  the pixel it was pressed at, and the control is found again afterwards by a
  key made of its tag, class and dataset — which eleven of the nineteen drawn
  menus there share, each of them changing its own text as it is chosen. So for
  the one control the reader had just used the tiebreak was empty, the fallback
  put a different menu at the pressed pixel, and the panel scrolled: measured,
  choosing a target moved it 994 px. Each control is named by the label beside
  it now, and the panel's menus are dressed before the correction is measured
  rather than a frame after it, which was worth a further 8 px on the first
  frame drawn.

Eighty-two checks drive the new picker on the real page with pointer events —
the drag and its snap, the bars, the ticks and the refusal to break an operon,
the end rules on both strands, the nudges, the junction previews, a real
genome-wide run and the design the names come out of — with six more in the
page suite for the arithmetic alone and five over the Target picker. Eight
reviewers went over it from different angles before it shipped, and what they
found is in the tests: twenty-nine reverts, one per decision, each naming its
own fault. One of them is a band that was never painted at all, because the
colour it asked for was a name this file does not define; another is a deletion
across the origin read as the stretch between its pieces, which is the trap this
project records in nine other places and now in a tenth.

Two limits worth knowing: the strip draws the first 160 genes of its window and
says how many it left out, and the ruler in the two cut figures runs its labels
together past a million bases.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.14

Changes in 1.1.50.14

Gene Studio 1.1.50.14

Three things asked for at the bench: changing an amino acid by pointing at the
amino acid, deleting an operon rather than one gene, and a workbook that stays
chosen.

• Point at the residue, not at the bases. Double-click an amino acid over
  the sequence — or take Change L2 · CTG from its right-click menu — and the
  mutagenesis window opens on that codon. Twenty residues grouped by what they
  are, and the stop; under each, the synonymous codons of that gene's own
  translation table, ranked by their share of the residue in the chosen host
  and marked where one is rare. Leucine in E. coli is offered as CTG 50 %,
  TTA 13 %, TTG 13 %, CTC 10 %, CTT 10 %, CTA 4 %, the last marked; glutamate
  as GAA 69 % against GAG 31 %, and changing the host to human turns that pair
  round without losing the codon already picked. Which codons are synonymous is
  a property of the table and not of the standard code — a gene annotated
  /transl_table=2 has four arginine codons where table 1 has six, and the
  window reads the gene's own.

• Picking one designs the inverse-PCR pair that puts it there. Not with a
  designer of its own: it asks the one the Guides view already uses and then
  chooses among the candidates that designer returns, so the two cannot come
  to disagree about what a good primer is. The window the guide arms are held to
  is the window here — annealing 18–50 nt aimed at 60 °C within 55–65 °C, a
  70 nt ceiling, a self-dimer screen at −9 kcal/mol, and a preference for an
  oligo that binds once — and the criteria are written under the pair rather
  than left to be taken on trust. Measured on a 3,666 bp construct, L2E came out
  as citA_L2E_F 22 nt / 60.4 °C and citA_L2E_R 30 nt / 60.4 °C, ΔTm 0.0 °C,
  with the new codon riding the forward primer's 5′ tail so the whole codon is
  replaced in one step. A reverse-strand gene is the part that is easy to get
  wrong by hand: the codon offered is the gene's own, and what goes into the
  molecule is its reverse complement (GCG → CGC). The mutant can be opened
  as its own construct, taken out as a GenBank file, or the two oligos copied as
  a row each.

• And it says what it will not do, with the reason. A linear molecule has no
  inverse PCR. A codon split by an exon junction or by the origin is named as
  its three separate bases rather than as a false span across the hole between
  them. A residue of the frame that runs on past a stop is not a residue of that
  gene. The codon already there is not a mutation. Where the record's own
  /translation disagrees with the bases, that is said rather than quietly
  resolved, and a rare codon is offered with its share named instead of being
  hidden.

• A deletion is sometimes one gene and sometimes an operon. The Guides view
  could delete the gene you picked and nothing else — an operon had to be
  deleted by reading two coordinates off the map and typing them in, which is
  the step that is silently wrong. There is a second box now, delete through
  this gene as well: an operon is a contiguous stretch by definition, so
  everything between the two genes goes with them, the way it is written on a
  bench sheet. A gene the stretch cuts through rather than removes whole is
  named next to it, because a partial gene left behind is the one thing nobody
  wants to find out about afterwards.

• A dropdown is not a way to find one gene among four thousand. The Target
  picker offered 4,153 unlabelled rows of name and coordinates on this machine's
  1.88 Mb record — reported from the bench as opening the list, picking one,
  going back and picking the next, because nothing on the row said which gene it
  was. It is typed into now: the name, the locus tag, the gene symbol, the
  feature type and the product are all searched, so division finds ftsZ,
  and each row carries the strand, the type and the product under the name. A
  searched list keeps every feature rather than being capped, and the coding
  features come first with their length in residues.

• A workbook chosen by hand stays chosen. It was forgotten the moment the
  window closed, so every launch went back to scanning a folder for
  *primer*.xlsx — a workbook kept anywhere else, or named anything else, was
  found by nobody, and the launch asked for it again once per launch, forever.
  The file is remembered and it wins over the folder, because it is the more
  specific answer; naming a folder goes back to scanning one; Forget the
  workbook is the other door. The remembered path names a file this app opens
  at launch, so it is the native side's to write and the page cannot set it —
  the capability check still stands in front of every read, but what this app
  reaches for on its own is not the page's to choose.

What holds the three in place: ten new tests in the page suite, a harness of
thirty-seven checks that drives the designer on the real page with pointer
events — the menu entry, the keyboard, the host change, the refusals and the
mutant it opens — forty-four more across the Target picker and the operon
stretch, and the dialog close-geometry audit extended to cover the new window at
six viewports. Two layout faults it found were in the shared dialog
chrome rather than in the new code, and they are fixed where the chrome is
defined.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.13

Changes in 1.1.50.13

Gene Studio 1.1.50.13

Fifty-two defects from a second adversarial bug hunt, across every half of the
application. The ones worth knowing about:

• A stop stays stopped. The licence record lives in the same preferences file
  the page writes its settings to, and a preference patch names keys — so a
  remove naming the desk's revocation took it out and the copy answered
  "licensed" again, with no new licence and nobody's authorisation. That key and
  the clock floor are the native side's now and the page cannot touch them.
  Three more doors into the same room are shut: an older signed notice replayed
  from the server no longer lifts a newer stop, pasting a stopped licence into
  the gate no longer unlocks it, and dismissing the gate the stop put up no
  longer leaves a stopped copy running — it asks about unsaved work and quits,
  and comes back if the quit is declined.

• A read with a bad patch in the middle is not a 98 % read. Scored as a
  mismatch, a run of unresolved calls cost more than a pair of gaps, so the
  aligner parked the Ns in insertions and invented four indels: 70 bases, 40 N
  and 90 bases against the sequence they came from answered 196 of 200
  matched, 98.0 % identity. An N is evidence of nothing and is now worth
  nothing: the same read answers 160 of 200, 80.0 %, no indel, while a real
  12 bp deletion and a real 9 bp insertion are unchanged. Copying a selection
  from a read that crosses the origin copies the calls picked rather than
  everything between the first and the last — 181 instead of 380 — and a
  deletion over the join is named in bases the molecule has.

• Three bases of a codon are one stretch, and a part found across the join is
  one part. A CDS at 13,209..5 reported its first codon as the whole plasmid:
  the start marker said starts at 1 and pressing it selected all 13,210 bases.
  A 10×His tag rotated over the origin was annotated as fifteen bases — five
  histidines of ten — and at 24 of a 150 bp fixture's rotations the whole match
  was dropped in favour of two shorter ones. Both are the same arithmetic, and
  the drawings, the terminator stems and the ribosome binding sites are read
  round the molecule with it: a binding site now answers the same at all 174
  rotations of its fixture.

• A record cut short is not a record with less in it. A GenBank file
  declaring 400 bp whose ORIGIN carries four opened as a four-base molecule, and
  a save would have written that out as a complete, shorter construct. It is
  refused by name now — measured against 7,715 GenBank files in this machine's
  own folders, not one of which trips it. A wrapped DEFINITION, an explicitly
  empty /replace, the spaces inside /PCR_primers, a .dna's creation date
  and sequence class, a blank row in a workbook and a location naming another
  accession all survive a round trip now; none of them did.

• A tail does not anneal, and a cached temperature is only the temperature of
  the conditions it was computed in. The facts line, the PCR advice and the
  Gibson table quoted the whole oligo's melting point beside a binding they had
  just said carried a 20 nt overhang — 75.4 °C where what pairs melts at 72.4,
  and 74.6 against 62.2 on a Gibson primer, so one too cold to work could not
  trip its own warning. And changing the salt or the primer concentration in
  Settings left every figure where it was: 72.1 °C and a 52 °C anneal held while
  500 mM Na⁺ makes them 84.1 and 64.

• A gene over a line break is pointed only where it ends. The linear map put
  an arrowhead on both pieces, the second of them claiming the gene stops at the
  end of the line; a spliced gene's connector hung 20 px under the exons it
  joins, and two exons within 5 px of each other were drawn at two heights. The
  navigator's cut marks no longer stay where they were when the picked enzyme
  changes, a thirty-character name is no longer written 124 px off the left of
  the window, and a go-to in the one-line sequence lands where the renderer
  reads it rather than 1.2 bases away.

• A referral is a code or the account they sign in with. The box took six
  characters with the capitals forced, so the e-mail somebody was told to put in
  it could not be typed — and what did arrive was stripped to six characters of
  nonsense that the desk then looked up as a code. Either shape goes through as
  it was typed now, and the desk resolves a code against the codes its holders
  have and an account against the people in it, saying by name when it resolves
  to nobody or to several.

Thirty-seven regression tests in the page suite and eight more across the
licence, save and packaging harnesses, each built from the measurement that
named the fault; the reasoning behind every threshold is in the project
record. One
finding went the other way: the record said a read's end is kept when it ties
the anchor, the code never did, and the code was right — nineteen calls agreeing
four of nineteen are noise, so the record was corrected instead.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.12

Changes in 1.1.50.12

Gene Studio 1.1.50.12

Golden Gate guide cloning, read round the circle and told what else the enzyme
does to the vector.

• A plasmid has no first base. Every read in the Golden Gate pair search was
  a flat slice of the sequence, so a stuffer sitting near a circular vector's
  origin had its window cut short and the pair the vector was built with went
  missing: No GGTCTC on the sense strand in the 10 bp before the stuffer,
  about a vector carrying that pair 2,990 bases away round the circle. The
  windows, the overhangs and the fill bases are all read across the join now,
  and the answer is the same whatever base the vector calls 1. The blind zone
  was a window's width at each end — 21 rotations of a 162 bp fixture and 21 of
  a 442 bp one, so it did not shrink as the vector grew.

• The enzyme's other cuts in the backbone are counted and named. A third
  cut anywhere else breaks the backbone into pieces no pair of oligos closes,
  which is the whole reason domestication exists, and nothing said how many
  cuts the enzyme has in the vector in front of you. The oligo panel and the
  design checklist now say so, with the size of the piece the nearest one
  leaves. Nothing is refused on it: 83–86 % of random 4 kb backbones carry
  another cut somewhere and a domesticated vector carries none, which is what
  makes the count worth printing when it is not. It counts cuts and not sites,
  because a site lying wholly inside the stuffer still cuts past its end.

• A blocked outer site is not a blocked flank. The pair is chosen outermost
  because in a digest every site cuts — but a host mark on that one leaves the
  inner ones cutting, and the arm ends at the first the host does not block.
  The template line now names that site and says what cutting there costs:
  more of the stuffer left in the product, or, where the cut only eats into the
  fill, nothing at all, in which case the oligos can be rebuilt rather than the
  template regrown. A flank that will still cut reads amber rather than red.

• Two cuts that pass each other leave nothing to ligate to. Reading round
  the circle made that geometry reachable, so it is refused: a circle has to
  keep some backbone duplex, and a linear template cannot be cut past its own
  end. A line's flush end — zero duplex, the top strand still there as an
  overhang — is an ordinary design and is not refused with them.

• Thirteen regression tests and the reasoning behind each threshold, from an
  adversarial review of the change.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.11

Changes in 1.1.50.11

Gene Studio 1.1.50.11 lets a guide be cloned with any Type IIS enzyme the picker offers, and shows a spacer whole.

• A Golden Gate enzyme is looked for as far out as it actually reaches. The pair of recognition sites that releases a vector's stuffer was looked for in 20 bases either side of it — the room a BsaI needs. An enzyme that cuts further out stands further back from its own cut, so on a vector built for it the site was outside the window and the design answered that the vector carried no pair. Of the 182 Type IIS enzymes the picker offers, 60 could not be used on any vector at all, and 65 more only where the stuffer was marked exactly on the cut. The window is the enzyme's own reach now.
• A recognition site is read as a pattern. 48 of those enzymes have an IUPAC code in their site — MmeI's TCCRAC, BaeI, BcgI — and were matched as literal text, so none of them was ever found on a real vector while a vector carrying the letters TCCRAC was. The site a design is built on and the site the digest draws are read the same way now. A site so unspecific that it is found in most vectors is refused rather than reported as a finding.
• Where a vector carries more than one site, the oligos are built on the outer one. Every site cuts, so the backbone ends at the outermost and the piece between comes away with the stuffer; a design built on the inner site cannot close the vector. Checked against this app's own digest.
• A pair that cannot replace what was marked says so. An enzyme whose cut falls inside the stuffer on both strands is refused with the number of bases in the sentence, as is one that cuts on both sides of its own site, and one whose two cuts between them leave nothing to replace. Where only the overhang reaches in, the oligos are still designed and the panel says how much of the stuffer stays in the product — the completed vector beside them replaces the whole of it, and those two had been disagreeing with nothing on screen admitting it. A vector that was chosen and could not be read now says why instead of falling back to a published convention in silence.
• A guide's spacer is read on one line. Genomic target context broke its rows every 24 bases while opening with 12 bases of flank, so a 20 nt spacer was split 12 and 8 across two lines — every guide, and the one thing the panel is for. A row ends where a region ends now, so the spacer and its PAM are read together, and the bases stay on the same column grid as the lines above and below.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.9

Changes in 1.1.50.9

Gene Studio 1.1.50.9 makes the 16S panel take the reads a sequencing service actually sends, shows them, and names the strain from NCBI, SILVA or JGI — with what BacDive knows about it.

• Forward and reverse reads pair wherever the primer sits in the name. Only the end of a name was read, so S1_27F_A01 and S1_1492R_B01 (a plate well after the primer), S1(27F), 27F_S1 and S1F/S1R each came in as two samples of one read, and nothing was merged. The primer is found anywhere in the name now, a well or run number after it is set aside, and S1F S2F S1R S2R is two samples rather than one. A word that merely ends in r, like Water, is no longer taken for a reverse read.
• Reads whose names say nothing are paired by their overlap. The forward and reverse reads of one amplicon overlap the wrong way round to each other, and two samples' forward reads never do; each read's best partner has to name it back. A read that overlaps two others about as well is left for you, with Pair with…, and a pair can be taken apart with Unpair. Where no name says which read runs along the gene, the conserved 16S primer sites decide, so the contig reads 5′ to 3′.
• The reads are on the panel. Reads lays both chromatograms under the merged sequence — the reverse read turned round, each peak on its own base — with every difference marked and stepped through, saying what each read called and at what quality. Each row names the files it was made from.
• .seq, FASTA and pasted reads are reads. A service's results folder can be dropped whole: the traces are taken, the .seq and _revComp.fasta copies of them are not read twice, and the PDFs and web pages are counted and left out. Paste reads takes a forward and a reverse read from anywhere.
• Identify against NCBI, SILVA or JGI. SILVA classifies a whole plate in one job on its ACT service, in about five minutes: SILVA's taxonomy to genus, GTDB's to species, and the nearest references with their identity. JGI BLASTs each sample against the 16S genes of IMG/M's isolate genomes, 10–40 seconds each, judged by species rather than by isolate. Neither offers an API, so the app makes the requests their own pages make; if one changes, the row says so and offers its page with the sequence copied.
• BacDive, when you want it. An identified sample shows what DSMZ's BacDive holds on the strain the search matched — type strain, culture collections, Gram stain, shape, motility, spores, oxygen, temperature, pH, salt, where it was isolated, biosafety level — and says plainly that it describes that strain, not your isolate.
• Submit sends. The homolog search no longer asks for a tick before it sends; what is sent, and where, is still written beside the button. A finished result opens at its top rather than below the fold.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.

Gene Studio 1.1.50.10

Changes in 1.1.50.10

Gene Studio 1.1.50.10 fixes 74 faults that ten outside reviewers found going through the app one area at a time, and keeps the window alive when reads are opened with text selected.

• Opening reads no longer stops the workspace display. Text was sometimes selected somewhere in the window when reads were chosen for the 16S panel: a construct name after ⌘A, or a line of the notes. About a second later the window went blank with The workspace display stopped, every time. The cause was the fade at the edge of a wide table. It was drawn by a scroll-driven animation, and Chromium cannot finish drawing a text selection in the frame where that animation falls out of date. The fade is still there, drawn another way, and the crash is gone.
• ⌘A selects in the Open panel. Picking a whole results folder used to take ⇧-clicks, because ⌘A did nothing in the system's Open and Save panels. ⌘V, ⌘X and ⌘Z did not reach their Go to folder field either. They all do now. In the window itself, ⌘A still selects the DNA or the text in a field, as before.
• "Identify every sample", not "every samples". The button that sends a whole plate to SILVA or JGI counts in English: every sample, the unanswered sample, or the 3 unanswered samples.
• A dropped trace stays where it was dropped. A .ab1 dropped onto a construct that already had reads vanished at the next tab switch. It is saved the way a read opened from the panel is, and it goes to the construct it was dropped on even if the tab changes while it is being read.
• An indel in a Sanger read is one event, where it is. A deletion just past the origin was counted as one-base deletions, one per base — near the end of a read, enough to cost the read its alignment. Bases a read carries past the end of a linear construct were placed before base 1. A read of Ns no longer aligns at 100 %.
• 16S: two samples stay two, and a plate is searched whole. Batch_S1F and Batch_S2R were merged into one sample, and a same-named read from another run could join the first run's. JGI was asked about the first 48 samples of a plate and the rest read returned nothing; a larger plate goes in runs of 48 now. A hit is not called settled past a species whose identity is unknown. A pasted read survives the panel redrawing, a second Identify no longer marks the first search failed, a SILVA job cut off by quitting can be sent again, and a reverse read's ambiguity codes show as their complements rather than N.
• A drag over the origin selects what was dragged over. On the map it selected past the end of the molecule, so a copy or a delete took only part of it, and a backwards drag on a linear map selected the wrong stretch. Cuts, introns and selections over the origin or a row break are drawn where they are: the navigator no longer paints the part left out, a single-row sequence paints both ends of a selection, a staggered cut over a row break is drawn on the right base, and a spliced feature over the origin keeps its intron. A search mark on a linear ring no longer runs ten degrees long, and hiding a feature takes it off the navigator.
• An edit keeps what it did not change. Changing bases inside a CDS left its /translation alone, so GenBank exported a protein the gene no longer makes; a translation that matched its bases is recomputed now, and one that never did is removed with a note. A case-only replace no longer merges a spliced feature's pieces, moving the origin keeps a whole-molecule feature whole and each piece's colour with it, and a megabase paste in mixed case goes in.
• Undo is each tab's own. Typing in a second tab after the first left nothing there to undo. After an undo the caret could sit past the end, where Backspace did nothing. And the History view put the saved mark on whatever version sat at the saved depth, not on the version that was saved.
• Files keep what they said. A circular join(1..10,91..100) is still read as crossing the origin, but it is written back as the file spelled it, and a location of three or more pieces keeps its order. A .dna keeps partial ends, order() and the <…> in a name or a note; one saved as GenBank keeps its accession, organism and references; and one cut short — a Dropbox file still syncing — says it is incomplete instead of opening with no features. A tailed primer keeps its /experiment and /db_xref, a long GenBank header is no longer cut at 64,000 characters, a GenBank record with no features no longer saves its sequence twice, and a circular FASTA comes back circular.
• A digest cuts where the enzymes cut. Two enzymes cutting at one place make one cut — EcoRI with SapI gave two whole molecules. An enzyme REBASE records no cut position for is listed as such, never drawn or digested as a blunt cut. Pin as lane works again. A short run keeps small fragments apart. A suggested digest is one that cuts this DNA as grown in its host, not an MboI that Dam blocks. One cut in a circle gives one fragment holding every feature. Importing SnapGene's enzyme sets no longer writes over one edited here, and a methylase mark on the bottom strand of a site is read as one.
• Primers and cloning. A Gibson product carries a donor feature's pieces to where they land. Golden Gate oligos for an enzyme that leaves a 3′ overhang put it on the right end. Inverse PCR refuses a linear template and says why. Removing the primer chosen in Amplify with no longer breaks the Primers view. A deletion's junction Tm is worked out in the order the primers make it. A sequencing plan for a selection over the origin plans the bases selected. The freezer finds a degenerate primer, and adding Gibson primers to a vector can be undone.
• Analyses. PGAP is given every contig of a draft genome and opens every record it returns. A locus tag two proteins share annotates neither of them. A failed inverse-PCR guide design exports no primers, rather than Golden Gate oligos under its name. Long ORFs and terminators are found through the origin, a terminator's span covers its T-run, and a run of Ns is not a palindrome.
• Search and History. A copy opened from History no longer shares its records, or its traces, with the construct it came from; editing it changed what an undo there would put back. A construct emptied of bases survives a restart with its history. A protein hit in a gene across the origin selects where it is, 2..1 on a circle selects the whole molecule, /…/i is a pattern again, and a pattern search that stops at 20,000 hits says so.
• Settings stay as they were set. A setting changed in the window no longer puts back what the app saved after launch — a renewed licence, the last folder, shortcuts — and preference writes land in the order they were made, quitting included.
• Safety. A stopped licence stays stopped however its signature is spelled. The Addgene sign-in hands only web links to the system. A PGAP result that cannot be copied is reported and kept rather than lost. A file reached through a chain of links is saved through the whole chain. On Windows and Linux a .faa opens from the command line, and on Linux a folder named in decomposed Hangul is opened as it is spelled.

What Gene Studio is, with pictures →

Given out one person at a time, for research. Not open source — the terms
are in LICENSE inside the application.

macOS: open the .dmg and drag Gene Studio to Applications.

Windows: run the setup .exe. If SmartScreen appears, choose
More info → Run anyway.

The app opens on a licence window and shows a Machine ID. Send that to
genestudio.help@gmail.com and paste back the licence you are given.