CRISPR Editing Outcome Record Checker
Last reviewed
BeginnerWhat you'll learn
~55 min- Ask an AI tool, in plan mode, for a program that reads the records a core's analysis software already produced and checks them under the core's numbered review rules
- Point it at the core's rules file, and read its plan for what matters: every rule by number, every expected sample accounted for, missing and unusable input kept apart, counts compared before rounding, supplied labels kept as they are, and a control concern beside every linked sample
- Check its list against findings worked out for the practice files, including missing records, ambiguous labels and an outcome exactly on the listing floor
- Say what a mark, a matching change and no additional mark can and cannot tell you, and who decides what can be reported
The story, and what you will end up with
It is Thursday 12 November 2026 at a made-up genome-editing core. You are the staff scientist preparing a review of recorded sequencing results for two orders. ORD-A asks for a sequence-level frame review of twelve cell-clone records, C-01 to C-12. Its order record says two copies of the site are expected. ORD-B asks for an inventory of outcomes in one founder record, F-01. Neither order asks this list to establish loss of function or inheritance.
The analysis has already happened. At this made-up core, you reconciled the orders, sample identities, plate record and controls against the source records, and checked which reference, coding annotation and analysis settings each result belongs to. The order record links C-01 to C-12 to CTRL-1 and F-01 to CTRL-2. The source results retain every outcome, including low-count and unresolved ones. The files for today’s review are as the analysis software and the typed sample list left them, and have not yet been checked against each other.
Your inputs are a sample list, a summary of counts and a table of outcomes with classifications and frame annotations already supplied. They stand in for an export assembled from source results and checked by an analyst. They are not a particular analysis program’s native export. Your own export may not supply a frame annotation for each outcome; check before you rely on that column.
Upstream analysis is a separate job. For example, the CRISPResso2 documentation, read 12 November 2026, says it “aligns sequencing reads to a reference sequence” and “quantifies insertions, mutations and deletions to determine whether a read is modified or unmodified by genome editing”. The program in this lesson reads classifications already recorded; it aligns nothing and assigns no classification or frame annotation.
By the end of this lesson you will have a small program on your own computer that reads these records, checks them under your core’s rules and lists which samples need a look, on screen and in a file your spreadsheet program opens. It makes no choice of clone or founder. You build it by asking your AI tool in plain English, and you check it against figures this lesson gives you.
🔍Words used in this lessonclick to expand
- Outcome: one distinct recorded change for a sample, such as
deletion 98-104orinsertion T after 102. Positions refer to an invented reference. They are descriptions to compare, not instructions for making a change. - Aligned reads: this practice export’s denominator. Its summary divides input reads into aligned and excluded reads. Its outcome counts include every aligned read, including reads classified unknown.
- Classification: the supplied label edited, unmodified or unknown. The program reads the label; it does not infer it from the change’s name.
- Frame annotation: a supplied label: frameshift, in_frame, unmodified, unknown, ambiguous or noncoding. Here in_frame means the source annotation records the reading frame as kept. A label is not a finding about function.
- Listed outcome: one with at least 5 percent of a sample’s aligned reads, this practice core’s own floor. Outcomes below it stay available, with their counts and annotations.
- Clone and founder: two kinds of sample record in these files. The rules for outcome count and frame review apply to the clone records only. That limited scope says nothing about a founder’s suitability or inheritance.
- Control association: the sample list’s explicit link to a control record. A control concern is shown beside every linked sample.
- Source-result pointer: an identifier for the original result. The practice files hold pointers, not the original analysis records.
- Checksum: a code worked out from a file’s contents, used here to identify the exact input version behind a list.
Before you start
First, if your AI tool is still open, leave it: type /exit and press Enter (if you use Antigravity, press Ctrl+D with nothing typed instead), and wait until you see the $ or % prompt. The next line goes there, in the terminal, not in the AI tool. It makes a folder for this lesson beside your practice folder and goes into it:
mkdir -p ~/bio-workbench/19-outcome-review && cd ~/bio-workbench/19-outcome-reviewStart your AI tool with the start line you kept from the setup lesson, the one with approval prompts off. Then ask it to set the folder up:
Set this folder up so we can undo any change you make, download https://learn.intellimetrics.net/starters/biotech/19/outcome-records.zip and unpack it into a data folder, and save a checkpoint.Look for it saying three things: the folder is set up with Git (the program that saves snapshots, from setup), four files are in the data folder (samples.csv, results.csv, outcomes.csv and review-rules.txt), and a first snapshot, your checkpoint, is saved. Later you can ask whether the files have changed since then. If it does not mention one of them, ask: “Did you save a checkpoint with the files in it?” It may also keep the zip file in the data folder, which is fine; if it says it kept one somewhere outside this folder, ask it to delete that copy. If the download does not work, download outcome-records.zip in your browser, check where it went (the browser’s list of downloads shows the folder), and tell your AI tool, for example: “outcome-records.zip is in my Downloads folder. Unpack that one file into a data folder here, and save a checkpoint.” In Windows PowerShell or a desktop app, “my Downloads folder” works as is; only say “my Windows Downloads folder” if you use Ubuntu (WSL) on Windows, which keeps its own Downloads folder apart from Windows.
The box below says what is in the practice files; later steps refer to them.
ℹWhat is in outcome-records.zipclick to expand
Three files, each of which opens in a spreadsheet program, and a rules file that opens in any text editor:
samples.csv: 15 expected samples: twelve clones, one founder and two controls. It records sample, order, batch, kind, reference, annotation, settings, expected copies, requested review, control association and source-result pointer.results.csv: 16 summary rows. Each holds the sample’s order, batch, reference, annotation, settings and source pointer, with input, aligned, excluded and unclassified read counts. A repeated row is kept; a missing record has no replacement row.outcomes.csv: 41 rows, each with sample, full change description, size, supplied classification, supplied frame annotation and reads. Unknown size is writtenunknown; the program never uses size to infer a frame.review-rules.txt: this practice core’s eleven numbered review rules, version 19.4, marked PRACTICE MATERIAL. They are this made-up core’s own, numbers included, not general quality standards.
Every row of the three spreadsheet files carries a practice marking. The reference, annotation, settings and source identifiers are invented pointers, not real sequences, software versions or analysis settings.
The rules file covers:
| Rule | What it asks the list to do |
|---|---|
| 1 | Identify missing, unexpected and repeated records; missing records give missing input, repeats give unusable input, and each needs a look. Keep every source row. |
| 2 | Mark invalid or uncheckable counts, zero denominators, contradictory values, missing fields and invalid labels as unusable input. |
| 3 | Divide outcome reads by aligned reads; list at 5 percent or more, using integer counts before rounding; retain every smaller outcome. |
| 4 | Mark positive whole aligned counts under 1,000, including unusable and unexpected input; zero belongs only to rule 2, and exactly 1,000 is not under the limit. |
| 5 | Mark more than two listed outcomes in an expected two-copy clone; give the single-outcome note only to clones and the scope note to founders. |
| 6 | Mark listed in_frame or unmodified outcomes in a clone’s frame review; do not apply this rule to founders. |
| 7 | Mark positive edited control counts before the floor and every sample linked to an unresolved control review. Other findings on the control keep their own rule numbers. |
| 8 | Compare clones’ listed changes within the same order, batch, reference, annotation and settings; use the full change, never size alone. |
| 9 | Mark unknown classification or unknown, ambiguous or noncoding frame annotations at any share; infer nothing. |
| 10 | Cite each fact once; keep input status and review marks together, suppress shares and outcome-set comparisons for missing or unusable input, and use no additional mark only for usable input with no concern. |
| 11 | Name the date, input files with checksums, row counts, source pointers and rule version behind the list. |
Rule 2 treats excluded reads as outside aligned reads, and unclassified reads as part of aligned reads. C-05 has 230 input reads: 212 aligned and 18 excluded. C-11 has 1,020 input reads: 1,000 aligned and 20 excluded; 100 of its aligned reads are classified unknown. No blank becomes zero.
The 5 percent floor is a listing rule, not evidence that smaller outcomes are absent. Every outcome remains in the file. The two-copy entry is the made-up order record’s premise; the read counts cannot verify it.
Everything in these files is made up for this course: the orders, samples, references, annotations, changes, counts and rules. No real analysis program produced them; no real gene, organism, strain, person or institution is represented. Each lesson in this track has its own practice files, with their own people and dates.
Switch to plan mode
Your AI tool is already open. Switch to plan mode now: Shift+Tab until it says plan, or /plan (the keys for each tool). In plan mode it usually reads your files, asks its questions and shows you a plan before it builds anything; the order can vary.
The conversation
Each message below is an example. Say it your own way.
Say what you want
Say what you do, which files to use, and what you want from them:
I review recorded editing outcomes at our core. Build a program I can run again that reads the files in data and lists what needs a look under our rules. Ask me what you need first.Your AI tool usually reads the files and comes back with questions, starting with your rules. It may already point out some of what it sees in the files. Leave its questions for now: send the three messages below first, then answer them. Your AI tool may wait for your answers or go straight on to propose a plan; either way, do not approve a plan until the three messages are sent.
Point it at your rules
These are this practice core’s own rules, in its rules file. Your core’s are in its own written procedures; later in this lesson you write them into a file of your own. Even if your AI tool has already read the rules file, tell it these are the rules to follow:
Our core's rules are in review-rules.txt in data. Follow every rule, and say beside each thing the list marks which rule it is for.Wait for its reply, then say what the program must do with the files:
Show the list on screen and in a file my spreadsheet opens. The program must never connect to the internet and must save only in folders I name.Wait for its reply, then say:
Leave my input files alone, keep every below-floor outcome available, and never write over an earlier list. Keep the program and checkpoints here; save practice lists in a reports folder here.Answer its questions
Answer each in a sentence. These are questions the tools often ask, with one way to answer each. Yours may ask them in a different order or show a list of choices:
| If it asks | You might say |
|---|---|
| Is a Python command from this folder all right? | Use whatever needs the least installed and works with no internet. |
| On screen, a saved report, or a spreadsheet file? | On screen and in a file my spreadsheet opens. |
| Which date goes at the top of the list? | The date I run it. |
| Should it stop or warn if the rules file changes? | Yes. Stop with a clear message if the rules file changes. |
| Does listed mean at or above the 5 percent floor, for usable input only? | Yes, that is how I mean it. |
| Is a repeated change the same change text twice, or the whole row? | The same sample with the same change text is a repeat. |
| Keep a size such as unknown exactly as written? | Keep the size as supplied and never use it. |
If it asks several questions at once, answer the ones about your rules and how outcomes are counted from your core’s own procedures, a sentence each, and accept its suggestions only about how the list looks. If you cannot tell which kind a question is, treat it as a rule. If it asks something not listed here, answer it the same way, in a sentence. If your core’s procedures do not settle a question, say you will check with whoever sets your core’s rules, such as the core manager, and do not guess a rule.
Read the plan
Next it usually shows you a plan: what it will build and how it will check it. Read it for six things:
- Every rule in the rules file is in the plan, by its number. Every expected sample is accounted for; repeated and unexpected records remain visible. Missing input and unusable input are distinct, and neither gets invented shares.
- Each share uses aligned reads, with numerator and denominator shown. Comparisons happen before rounding, exactly 5 percent is listed, and below-floor outcomes stay available. Excluded and unclassified counts keep their different meanings.
- Annotations stay as supplied, including ambiguous ones. Clone rules and the single-outcome note apply only to clones; the founder scope note is present from the start.
- Control checks read every classified outcome before the floor, and an unresolved concern makes every associated sample need a look under rule 7, even when no other rule marks it or its input is missing or unusable.
- Matching outcomes use the full recorded changes within the same context, never sizes alone. Every reason names its rule; nothing establishes origin, function or inheritance.
- It is a program you can run again, with the date, input identities, counts, source pointers and rules on each list. It never connects to the internet and saves only in folders you name: the program and checkpoints here, practice lists in reports here. It leaves your files as they are and never overwrites an earlier list.
A first plan often leaves one of these out, even when you have asked for it. Say so in a sentence, in your own words, for example:
The plan doesn't show rule 7 beside samples with missing input. Keep the unresolved control concern beside every linked sample, including those.When the plan has all six, approve it: if your AI tool offers a choice to go ahead, pick it; otherwise say “Yes, go ahead and build it.” In Codex, press Shift+Tab first to leave plan mode. In Copilot, saying it brings up a choice to go ahead; pick that. Let it build. It may run what it built and tell you how to run it yourself; either way, the next step asks it to. When it has finished, ask it: “Did you install anything, and where?” Nothing, or something inside this lesson’s folder, is fine for these lessons. If it names anywhere else on your computer, such as the zip file it downloaded, ask it what that is for, and to remove it if you do not want it there.
Look at what it built
Ask it to run what it built:
Run it on the files in data and show me the list.Your list will be worded and laid out in its own way, and some details may be only in the spreadsheet file, not on screen. Each mark should name its rule by number. Check it for these:
- The header under rule 11 names the date, four input files with checksums, rules version 19.4 and rows read: 15 sample rows, 16 summary rows and 41 outcome rows. All rows can be read as text; some contain invalid values. Those rows must still be listed with their reasons.
- C-09 is missing its summary and outcomes, giving missing input and needs a look under rule 1 only for those facts. Its separate linked-control concern is under rule 7. X-01 is unexpected under rule 1. C-12 has two summary rows, giving unusable input and needs a look under rule 1; neither is chosen. X-01’s input status is unusable input only, not missing input: it has a summary row and an outcome row, and missing input is for expected samples. If your list also calls X-01 missing input, say: “X-01 is labelled missing input, but it has records: it is an unexpected sample with unusable counts. Label it unusable input only.”
- C-12 also has unusable input under rule 2 for different facts: its summaries disagree on reference and aligned count, one reference conflicts with the sample list, one row fails input-count reconciliation, and its 900 outcome reads match neither 1,000 nor 999 aligned reads. X-01 is unusable under rule 2: zero aligned reads and an outcome count of minus 1, so its outcome total cannot be checked. Neither gets a percentage. C-12 also needs a look under rule 4 for its second summary’s 999 aligned reads. X-01’s zero denominator is cited only under rule 2, with no duplicate depth mark under rule 4.
- C-02 needs a look under rule 6: its in_frame outcome has 4,733 of 9,904 aligned reads, 47.79 percent.
- C-03 needs a look under rule 5: three listed outcomes, with 5,702, 3,841 and 2,690 reads out of 12,377.
- C-05 needs a look under rule 4: 212 aligned reads, below 1,000. Its shares, if displayed, say depth concern; its 18 excluded reads are shown separately.
- C-06 and C-08 need a look under rule 8. Both list
deletion 101-102andinsertion T after 102. C-07 listsdeletion 99-100andinsertion A after 102, the same sizes but different changes; it gets no rule 8 mark. - CTRL-1 needs a look under rule 7: 28 edited reads out of 8,940, 0.31 percent. Both edited outcomes are below the floor. All twelve expected clones, C-01 to C-12, also need a look under rule 7, with CTRL-1’s unresolved concern as the reason, including missing C-09 and unusable C-12.
- C-10 lists its in_frame outcome under rule 3 at exactly 50 of 1,000 reads, 5.00 percent, and needs a look under rule 6 for that listed frame annotation in its requested review. Its depth is exactly 1,000, so rule 4 does not mark it.
- C-11 needs a look under rule 9: 100 of 1,000 aligned reads have unknown classification and ambiguous frame annotation. Its 20 excluded reads are outside that denominator.
- C-01, C-04 and C-07 each say needs a look under rule 7, with CTRL-1’s unresolved concern as the only reason; none says no additional mark. C-01’s one listed outcome is 11,690 of 11,842 reads, 98.72 percent, with rule 5’s note that the table cannot establish that every copy was represented. C-04’s in_frame outcome remains available: 131 of 10,615 reads, 1.23 percent, below the floor.
- F-01 has four listed outcomes out of 13,460 aligned reads: 4,980, 2,961, 2,288 and 1,884 reads. The last is unmodified, 14.00 percent. Its under-floor subtotal is 1,347 reads, 10.01 percent, including an in_frame outcome with 619 reads, 4.60 percent. F-01 says no additional mark under these rules, with no clone-rule mark or single-outcome note; its scope note says those rules do not apply and makes no inheritance conclusion.
- CTRL-2 has 7,605 unmodified reads, zero edited reads and no additional mark. Neither control gets rule 5’s single-outcome note, although each has one listed outcome. CTRL-2’s linked founder has no control concern under rule 7. No additional mark is not permission to report.
- Every sample retains its source-result pointer and source row numbers under rule 11. No sample is labelled confirmed, correct, good, best, safe, passed or approved.
- When you ask “Have my record files changed since the checkpoint?”, it compares them with the checkpoint and says no.
- Look at the line it used to run the program and ask: “Does the program connect to the internet, or save anything outside the folders in that line?” Look for no to both, with the program and checkpoints here and practice lists in reports here.
Percentages above are rounded to two decimal places for reading. Rule 3 owns listing at or above the floor and retaining each smaller outcome, with no review mark for either fact alone. The floor uses the integer counts, not a rounded percentage. Under-floor totals are separate from listed outcomes; nothing is rescaled to make a shorter list add to 100 percent.
If something on the list is not true, tell your AI tool in a sentence what you see and what you expected, for example: “C-01 has no control concern. Rule 7 says CTRL-1’s unresolved concern appears beside every linked sample. Add it.” If a line or a reply calls a sample, clone or founder confirmed, correct, good, best, safe, passed, a knockout or fit to report, or ranks or chooses one, say: “Take that off. Show only the recorded counts, input status, rule numbers and reasons.”
Notice what matching changes leave open
CTRL-1’s edited reads stay available under rule 3 and its positive edited count has a mark under rule 7. Ask your AI tool:
Show CTRL-1's edited changes below the floor and the usable ORD-A samples recording each same change. What can this table establish about the cause? Don't change anything.Before you read on: does a matching change tell you where those reads came from?
CTRL-1 has 19 reads recorded as deletion 98-104, also recorded in C-02 and C-03. Its other 9 edited reads are deletion 102, also recorded in C-01, C-04, C-05, C-10 and C-11 among usable ORD-A inputs. The raw outcome rows of C-12 and X-01 name that change too, but both are unusable input and stay out of this comparison. F-01 belongs to another order and reference, so matching text there is outside the question.
Those are matches between recorded descriptions. They add no rule 8 mark: that rule compares listed sets between clones, not a control’s changes with a clone’s. The table has no evidence that establishes a cause, a direction or a source for the control’s reads. Follow the source-result pointers back to the original records and ask what those records support. The control review stays unresolved for the associated order until the staff scientist handles it under the core’s procedures; the list does not decide which results can be reported.
Read what each mark and note covers. C-01 needs a look under rule 7 because CTRL-1’s review is unresolved, even though it has no other mark. Its rule 5 note applies only to clones: a single listed outcome is not an inventory of every copy. C-04’s small in_frame outcome did not disappear when it fell below 5 percent. F-01’s four listed outcomes do not receive the clone checks at all, and its 14 percent unmodified share and below-floor outcomes still need reading. A person’s decision cannot add evidence the records lack.
The list checks recorded counts against the core’s rules; the staff scientist and ordering lab decide what the evidence supports and what can be reported under the core’s procedures.
One real run
Here is part of what one AI tool said after the question above, with Codex on 12 November 2026. Yours will word it differently; check your list against the checks above, not against this.
**Rule 7:** All **28 edited control reads** count before applyingthe floor, so CTRL-1’s concern remains unresolved.
The table establishes that these exact change descriptions occurin the control and the listed samples. **It cannot establish thecause, common origin, or direction of transfer.** The overlapalone does not demonstrate contamination or identify a source.The samples, changes and counts it talks about are made up for this course.
How real records differ
- Check how your export defines each denominator, which reads it excludes, and whether it retains unknown classifications and every small outcome. Compare its totals with its original analysis before using a floor.
- Check the reference, coding annotation, analysis version and settings behind each classification. One program’s documentation says: “Mutations within this number of bp from the quantification window center are used in classifying reads as modified or unmodified.” (CRISPResso2 parameter documentation, read 12 November 2026.) That describes that program’s setting; it is not an instruction to choose one. Check the documentation for your own version and what was recorded for your analysis.
- Check where each per-outcome frame annotation came from. An aggregate count does not identify which annotation belongs to which outcome in your export. If your records do not supply that relationship, this program cannot invent it.
- Check the evidence behind any expected-copy entry, and what your source records establish about copies represented by the assay. Do not turn a single listed change or a read proportion into a copy count.
- Check whether two descriptions mean the same change on the same reference before comparing sets. If your source uses inconsistent descriptions, have the analyst settle the mapping; matching text alone is not a decision about sample origin.
- Check your order identifiers, batch boundaries, control associations and the record that holds control review. Check which person your procedures name to settle unresolved concerns, and where the reporting decision is recorded.
- Check what evidence the ordering lab actually requested and what the records support about function or inheritance. The practice table carries neither kind of evidence.
Keep going
Small changes do not need plan mode. Say each in a sentence, then run it again:
- “Make a printable sheet for each sample, keeping its control concern, every outcome and the list’s header.”
- “Does the program name any sample from these practice files? If it does, make it work those out from the files instead.” A program built in a conversation may keep your examples written into it, and on the next order’s files it would get them wrong.
- “Save a checkpoint.” Your AI tool saves a snapshot you can return to.
A change to what the list counts or marks is not a small change: it is a change of rules, made as in step 4 below, with the made-up examples checked again before the program is used on real files.
When you want to check a real order’s outcome records, use the finished program. It runs on your real files by itself, with your AI tool closed, so your AI tool does not see them; step 7 keeps it out of that folder afterwards too. Nothing confidential or unpublished, and nothing about people or about animals under a protocol, goes to your AI tool or into this course folder. A real order’s records never go to your AI tool: not a row, not a value, not a screenshot of an error. Whether they may go anywhere else is set by your core’s policy and any agreement that covers the work. Your inputs may come from an approved export assembled from your core’s analysis results and order records. Your core’s export may not be shaped like the practice files; check its columns, and what each one means, against theirs. Do these steps in order; the first five happen now, while your AI tool is open and this folder still holds only the practice files:
-
Go through the data check, the questions in setup about what the data is, who decides how it is used, and which AI tool would see it. Go on only if the people who decide allow this use.
-
Decide on a folder of its own for your real files, outside this course folder, for example
~/outcome-records. Nothing has to be in it yet. If the folder is not there, ask your AI tool to make it: “Make an empty folder called outcome-records in my home folder.” -
Ask your AI tool for blank files to fill in, and to show the program working on them:
Make fill-in copies of the three data files in ~/outcome-records, with one made-up sample and its control, that my spreadsheet opens. Keep the program and checkpoints here. Run it on them and show me.Look for it accounting for both made-up samples and their outcomes. Your rules file is not there yet, so this list may use the practice rules or say that no rules file was found; step 4 gives it your core’s rules. The real-files folder holds data and the rules file, not a copy of the program or its checkpoints. If your core’s export has other column names, tell your AI tool those names and what each column holds, in general words, with two rows you make up in the same shape, made-up IDs and counts, never real rows. Share only the column names and the layout; if your core treats them as confidential, ask whoever is responsible first, or make up equivalents. Take what each column holds from documentation your core lets you share, not from its heading alone. Then ask: “Change the program to read files like these made-up rows.”
-
Have your AI tool make a rules file for your core’s own rules, while only made-up data is there:
Make a plain-text rules file in ~/outcome-records, named the way the program looks for one, with one placeholder line. Tell me its name.A rules file holds rules, not records: no order, sample or person from a real order goes in it. Your AI tool reads the rules file, so if your core treats its procedures as confidential, ask whoever owns them before any of their wording goes to it. Then open that file in any text editor and replace the placeholder line with your core’s own review rules, from its written procedures, the way the practice file has them: numbered, one rule to a paragraph, in ordinary sentences, with their version. Write only rules that compare recorded values; anything that needs the reads themselves, an alignment or a classification stays with your core’s own validated software. Word each rule so only one reading fits. Save it under the same name, and tell your AI tool:
Our rules are in that file now. Change the program to follow those rules instead of the practice ones, and update the blank files to match.It may ask how your rules apply, such as which columns to keep. Answer each from your core’s own procedures, a sentence each, and accept its suggestions only about how the list looks. If you cannot answer one yet, say so and ask it to wait; do not let it go ahead on a rule of its own choosing. If it does not ask, ask it how it read each rule. Wait for its reply, then:
Make up one sample for each of our rules and show me the list, including a control concern, unusable input and a value exactly on each limit.Where your rules overlap, a made-up sample may be marked by more than one, and the tool should say every rule that marks it.
Before you read its list, note what each made-up sample should get under your rules. Then check that each made-up sample needs a look under the rule it was made for, with that rule’s number and reason, and that the control concern, the unusable input and each value on a limit come out as your rules say. If a limit has no example just below and just above it as well, ask for them. If a rule compares samples with each other, ask for two that should match and one that should not. The program follows the rules it was built for, and your AI tool may build it to stop when the rules file no longer matches them. If your procedures change a rule, change your rules file, ask your AI tool to change the program to match while only made-up files are in your folder, and check it the same way. Then ask: “Save a checkpoint.”
-
Ask your AI tool for the exact line:
Write me the exact line to run this on the files in a folder called outcome-records in my home folder, with full paths, saving everything it writes in a new folder called outcome-lists beside it, never in this folder.Use your own folder’s name. If it offers to copy your real files into this folder instead, say no. Copy the line it gives you into a note. It will look something like this one, though yours may differ:
Terminal window mkdir -p /home/dana/outcome-lists && python3 -B /home/dana/bio-workbench/19-outcome-review/review.py /home/dana/outcome-records --out-dir /home/dana/outcome-listsRead it left to right.
mkdir -pand the folder after it make the lists folder if it is not there yet, and&&runs the rest only if that worked.python3runs a Python program, and-Bstops Python leaving its own cache files beside the program. Next comes the checking program, in this lesson’s folder; then the folder holding the three data files and your rules file, which the program finds there by name;--out-dirand the folder after it say where the list goes, a new file each run. Your line may not have themkdirpart if the folder is already there. If your line sends the lists anywhere else, such as inside outcome-records or the folder that holds it, ask: “Send the lists to a new folder called outcome-lists beside outcome-records, and give me the line again.” Use the new line. The line depends on the computer: the one above came from a Linux computer, so use the line your own AI tool gives you./home/danais the home folder of someone whose user name is dana; your own user name goes wheredanais, and on a Mac the folders begin/Users/instead of/home/. The line runs the program where it is, so leave this lesson’s folder in place. Then ask your AI tool what the program needs in that folder: “What must the three data files and the rules file be named, and which columns must each data file keep?” Note its answer beside the line. Then ask: “Does the program connect to the internet, or save anything outside the folders in that line?” Look for it answering no to both. Its answer describes the program; it is not a test: the trial below shows where files land, and if your core needs proof that a program makes no network connection, ask your IT staff before real use. If you are unsure about any part of the line, including where any extra output is saved, ask your AI tool now, while it is still open.Then, before you leave your AI tool, try the line on the made-up files, because a real order’s records must never end up in the course folder:
Run that line on the made-up files in ~/outcome-records. Where did everything it wrote land, and did running it change anything in this lesson's folder?Look for every output in the folder your line names, nothing changed in this lesson’s folder, and an earlier list left untouched. If anything landed in this lesson’s folder, tell your AI tool so and ask it for the line again. Before leaving the tool, ask it to remove the made-up rows, any made-up examples it put in a folder of their own, and every list made from them, so that outcome-records holds only the blank data files and your rules file.
-
Leave your AI tool the way you did at the start of this lesson.
-
Put your real export files in outcome-records under the names you noted, in place of the blank data files and beside your rules file. If you fill in a blank sheet, use the source records your procedures name and check the entered values. As setup says, a tool started with approval prompts off can open other files your account can open, so from now on start your AI tool only in your course folder and never ask it about the outcome-records folder. Paste the line from your note at the
$or%prompt and press Enter. The list appears in the folder your line names, not in the course folder. To come back to the lesson folder later, type this lesson’s folder line again, the line near the top of this page that begins withmkdir. -
Before you rely on the list for a new order, run it on a past order whose records you have already reviewed yourself, and compare every sample it lists, input status and reasons included, with what you found. If they differ, find out why first, against the original records. That comparison is about that order, not an approval for later ones: use no list for reporting until the program has been accepted the way your core’s procedures say, by whoever they name. Reproduce a problem with made-up records before you ask your AI tool about it.
Try another case
Before the quiz, write two or three sentences for yourself. A colleague sees C-01’s 98.72 percent and asks to report it immediately, overlooking its needs-a-look mark under rule 7. What does the list show, what does it leave open, and who decides?
A good answer names CTRL-1’s unresolved concern as C-01’s reason for needing a look, reads the clone-only single-outcome caveat, and goes back to the source records. It does not turn a read share into evidence about every copy or function. The staff scientist and ordering lab decide what can be reported under the core’s procedures from the evidence available.
Key takeaways
- A short ask, then your rules file. Say what you do and which files; point it at your core’s numbered review rules, written from its own procedures, before you read the plan.
- Read the plan for every rule by number: every expected sample accounted for, missing and unusable input kept apart, counts compared before rounding against a named denominator, below-floor outcomes kept, and an unresolved control concern beside every linked sample, even one with no other mark.
- Check a new list against findings you know before you trust it on an order you do not.
- Matching changes are recorded similarities, not a finding about cause or origin.
- A mark is a recorded value that needs a look, and no additional mark is not permission to report. The staff scientist and ordering lab decide what the evidence supports and what can be reported, under your core’s procedures.
- Practice data stays practice data: real records go to the finished program, with your AI tool closed, after the data check.
CTRL-1 has 28 edited reads out of 8,940 aligned reads, all below the practice core's 5 percent floor. C-01 has no other reason for a mark. What should the list show?
What’s next
The next lesson is about reviewing the protein groups in a finished search’s results under a core’s reporting rules.