Keep the useful allele. Find out what came with it.
A plant breeder can recover most of a familiar variety and still carry a troublesome stretch of DNA next to the one allele they wanted. This guide shows how to spot that local problem, how to read a pair of flanking markers, and why a tidy marker result still needs a field check.
Imagine a pea line that holds up well in a wet season but matures later than buyers need. A grower’s usual line matures on schedule but lacks the particular resistance allele a breeder wants to introduce. The objective is precise: keep the useful allele, recover the familiar background, and find out whether delayed maturity came along for the ride. Those are three questions, and one test cannot answer all three.
All plant names, data and counts in the worked cases are invented. You can do the reasoning with slips of paper and a pencil. No breeding facilities or personal genetic information are needed.
Genetics subject guide · Learning that transfers into life · How X Works hub
Choose your reading route
Route 1 · Chapters 1–3: Start with the chromosome
Begin with two parents and one clear promise
Understand why a good-looking background percentage can hide a local problem.
Route 2 · Chapters 4–6: Follow the evidence through a cross
Work through an Orchard backcross one offspring at a time
Choose informative offspring, work through the counts and check what the markers can establish.
Route 3 · Chapters 7–9: Make a claim you could defend
Check whether the plant kept its promise
Confirm the plant result and practise transferring the method when the geometry changes.
Full chapter index · Independent worked case · Questions and practical checks
Chapter index
Start with the chromosome
Begin with two parents and one clear promise
See why the target has neighbours
Give each marker a job
CHAPTER 1 OF 9 · Start with the chromosome
1. Begin with two parents and one clear promise
Start with a recipient line called Orchard. It has the harvest timing and pod quality a grower already wants. Ridge is the donor. It carries a particular resistance allele, which we will call R. Orchard carries r at the same locus. Neither letter tells us whether the trait is dominant. They are names for two versions of a locus, not a ranking. We will assume for the first example that a validated DNA assay can distinguish R from r in these two parents.
Cross Orchard and Ridge. The first-generation plant carries one copy of each version at the target. Backcross that plant to Orchard. In a simple diploid model, the Orchard parent supplies an r copy. The other parent may supply R or r. If we need R to continue the project, the offspring that inherited r from both parents are not useful for that particular branch, however much they resemble Orchard. Checking the target first prevents a very efficient programme for recovering the wrong plant.
Orchard is called the recurrent parent because the selected offspring are crossed back to it in successive generations. Ridge is not crossed in again on each round. In the simple model, with no selection and equal parental contributions, expected recurrent ancestry is one half in the first cross and three quarters after one backcross. On the next backcross it is seven eighths. These are averages for the unselected model, not guarantees for any seed, and not an appropriate calculation of ancestry immediately around a deliberately retained donor allele.
Before looking at results, write three statements the project would have to support. First: this plant has the intended allele. Second: these observations limit how much linked donor DNA remains on each side. Third: under stated trial conditions, the plant retains the useful resistance and meets the harvest criterion. You can have strong evidence for the first and weak evidence for the other two. Keeping the statements separate tells you what to measure next rather than making every disappointing result look like a failed assay.
CHAPTER 2 OF 9 · Start with the chromosome
2. See why the target has neighbours
Sketch one long line for each chromosome copy. Put a dot labelled R on the Ridge copy and r at the matching position on the Orchard copy. Add two more dots, one on either side. They stand for DNA positions where Ridge and Orchard differ. Now shade the Ridge chromosome. If an offspring receives that whole shaded stretch, it receives every allele there, including any that affect maturity. This is why selecting the right allele is not the same as selecting only that allele.
During meiosis, paired homologous chromosomes can exchange corresponding material. On your sketch, cut each copy between the left marker and the target and swap the two right-hand pieces. One product now has an Orchard left marker but retains the Ridge target. That product matters because a breeder can select the useful change without selecting away the desired target. If the cut were on the other side of R, it would instead help resolve the right flank. Location matters more than how dramatic the change looks in an ancestry report.
Linkage drag is the unwanted effect of retaining donor material linked to the desired allele. A retained donor segment is a risk for drag, not automatic proof of harm. Some neighbouring donor variants may have no meaningful effect; some may even help. Conversely, if R itself changes more than one trait, narrowing the neighbouring segment will not necessarily separate those effects. The chromosome sketch tells us how to test a neighbouring cause. It does not tell us that every disappointing phenotype has one.
The natural temptation is to keep crossing back and wait. More generations create more opportunities for recombination, but they do not certify that a useful breakpoint has occurred. The selected target continues to favour its neighbourhood until suitable offspring are found. Decide whether the immediate task is reducing a linked donor interval or recovering the wider Orchard background. Both are legitimate; measuring one and reporting it as the other is the mistake.
CHAPTER 3 OF 9 · Start with the chromosome
3. Give each marker a job
A foreground assay asks whether the offspring carries the intended target allele. In the example it reads the R/r position directly. A marker linked to R is sometimes used instead, but then the proxy can become separated from the target. If you only have a proxy, state what validates it for these parents and what extra evidence you will need before claiming the allele is present. A marker name on a report does not tell you what was measured.
A pair of flanking markers asks a different question. In the order Left–Target–Right, each outside marker is informative only if Ridge and Orchard can be distinguished there. Suppose the Ridge pattern is D–R–D and Orchard is O–r–O, where D and O mean the parental version of that marker. If both parents have the same version at Left, a match to Orchard at that position tells you nothing about ancestry. Find an informative marker, or keep the left side unresolved.
Background markers sample other parts of the genome. They are useful for comparing candidates with similar local breakpoints, or for checking that a small set of candidates has not retained unexpected donor blocks elsewhere. A reported percentage depends on which positions were tested and how the score was calculated. For example, counting the proportion of markers at which both copies are Orchard is different from counting the fraction of all observed allele copies that are Orchard. Neither is a base-by-base measurement.
Keep the jobs on separate lines of the record: target, left interval, right interval and background. Use a fifth line for phenotype. A decision becomes easier to audit when each measurement has a question beside it. If the left marker fails because the two parent controls both produce the same unexpected band, the honest entry is unresolved, not Orchard. If the target assay passes but the candidate matures late, keep the target result and investigate the trait rather than rewriting the genotype.
CHAPTER 4 OF 9 · Follow the evidence through a cross
4. Work through an Orchard backcross one offspring at a time
Imagine four offspring that have all tested R/r at the target. Their other chromosome came from Orchard. For the chromosome carrying R, the informative patterns are Cedar D–R–D, Birch O–R–D, Elm D–R–O, and Larch O–R–O. In a real lab these patterns have to be established from the pedigree and informative genotypes; here the phase is given. Read the middle first, then inspect the left and right positions separately.
Cedar retains donor marker alleles at both ends. That is compatible with retaining a long donor stretch. Birch shows that the chromosome carries Orchard ancestry at Left but Ridge at Target, so a transition is inferred between those positions in the simple inherited-segment model. Elm gives the corresponding evidence on the right. Larch puts Orchard alleles at both flank markers while keeping R. It is the best local candidate among the four under these assumptions. It need not be the best candidate for every breeding objective.
Suppose a broader marker panel gives Cedar the highest Orchard score. If delayed maturity is suspected to come from very near R, Cedar is not automatically the best choice: its score mostly speaks to sampled positions away from the target. If Larch has a substantial donor block elsewhere, do not hide it either. Keep Larch for the local question, perhaps keep a second line for comparison, and work on the background with the evidence that speaks to it. The choices are easier to explain if you preserve the rejected options and the reason for rejecting them.
Now imagine a new informative marker halfway between Left and Target. Larch gives the donor version at this new position. That does not contradict the original result. It narrows where the left ancestry change could lie, between the original Left and the new marker in the simple model. The retained segment on that side still includes the new marker. This is a good correction to write in full: the outside marker passed, but the inside marker reveals donor ancestry closer to the target. Neither marker tells us the exact breakpoint.
CHAPTER 5 OF 9 · Follow the evidence through a cross
5. Translate a recombination chance into a screening question
Suppose a teaching cross gives a conditional probability of 0.04 that a target-positive offspring has the desired left-flank recombinant pattern. We condition on target-positive offspring because those are the candidates screened for the next question. In 30 independent eligible offspring, the chance of missing the pattern every time is 0.96 to the power of 30, about 0.294. The chance of finding at least one is therefore about 0.706. The expected count is 1.2, but that does not mean the smallest possible result is one.
If the goal is at least a 95% chance of finding one, the same simplified calculation asks for the smallest whole number n with 0.96 to the power of n at most 0.05. That is 74 eligible offspring. At 73 the chance falls just short; at 74 it just passes. This is not a recommendation to sow 74 seeds. If roughly half of the surviving offspring are eligible at the target, 74 eligible would mean about 148 surviving offspring in expectation before other losses. A fixed starting number does not guarantee the required eligible count.
What if a colleague suggests multiplying the left and right rates to estimate a two-sided recombinant in the same offspring? Ask first whether the two events are reasonably independent in the relevant meioses. Crossover interference, interval structure, viability and selection can all make the simple multiplication misleading. A breeder may instead keep a validated one-sided recombinant and seek the second breakpoint in a later backcross. Then the left side should be checked again rather than assumed to stay correct.
The count is a design check, not a promise. The 4% came from a fictional cross. In a real cross the estimate has uncertainty, and the same physical distance need not imply the same recombination frequency in a different family or chromosome region. Record how many plants germinated, how many produced usable assay results and how many were target-positive. Otherwise a poor assay or strong survival bias can masquerade as an unusually low recombination rate.
CHAPTER 6 OF 9 · Follow the evidence through a cross
6. Draw what is known between markers, not what you hope is there
Place Left at position 10, Target at 14, and Right at 19 on an imaginary physical map. If the target-bearing chromosome reads O–R–O, you know the origin at the two informative markers and at the target. With one ancestry transition on each side and a single retained donor tract around R, the left transition lies between 10 and 14 and the right between 14 and 19. Those bounds belong to that model. You cannot report a precise start or end coordinate just because the positions of the markers are precise.
If an inside marker at 12 is D and an inside marker at 16 is O, the simple intervals become 10–12 for the left boundary and 14–16 for the right. A picture that labels donor ancestry from exactly 12 to exactly 14 would be wrong: those are known donor positions, not the full segment. The segment extends at least between those donor positions in the single-tract model, and may extend beyond them up to the next limits. This distinction matters when a candidate maturity variant is at position 11.
A marker report may also miss short tracts between assay sites. Multiple ancestry transitions, genotyping error or structural differences between parents can make the one-transition picture too simple. If your conclusion depends on the ancestry of one particular position, design a test at or near that position and state its limitations. If your next decision is only which plant to keep for another generation, the rough boundary may be good enough, provided the uncertainty goes forward with the plant.
Map units deserve their own line. Physical distance counts bases; genetic distance describes recombination on a map. A short interval in bases can still be awkward to separate in practice. Likewise, moving a flank marker farther away may make a recombinant easier to find but leave a wider interval in which unwanted donor material can hide. The best marker position depends on the decision: initial screen, precise interval check or later confirmation.
CHAPTER 7 OF 9 · Make a claim you could defend
7. Check whether the plant kept its promise
Suppose Larch has the desired target and the shortest inferred segment. You grow Larch, Orchard and Ridge together in a small, properly identified trial. Orchard and Larch mature near the same day, and Larch appears more resistant in the test conditions. This supports a result for that trial, not a universal promise. Ask how maturity was defined, how resistance was assessed, whether plots were comparable, and how much variation there was among repeats.
Imagine instead that Larch still matures late. Do not conclude at once that the flank markers failed. The donor segment remaining between them might still contain the relevant variant. A different retained donor region might matter. The target allele could have an effect on both traits, or the plant’s response could depend on the genetic background or growing conditions. Sample handling and scoring errors are also possible. These explanations predict different evidence, which is why keeping more than one independently recovered line can help.
For example, if several lines retain R and have different breakpoints, compare their maturity while accounting for differences elsewhere. If lateness follows a shared segment rather than the mere presence of R, that suggests a linked cause worth testing. If every otherwise comparable R line remains late despite substantially different linked intervals, investigate effects of R itself and other shared factors. Neither pattern alone proves mechanism, especially if the lines still share a hidden donor tract or were all tested under one stressful condition.
Do not collapse phenotype confirmation into a single average if the claim is about dependable timing. A mean can hide a broad or uneven response. Report the comparison conditions, spread and number of independent plots or units, and what changed relative to the recipient. Seed identity and repeated genotype checks should travel with trial records. Otherwise it becomes impossible to tell whether a later failure reflects biology, a changed environment or a different seed lot.
CHAPTER 8 OF 9 · Make a claim you could defend
8. Keep a record another person can use
Put a single candidate ID on the plant, leaf sample, assay sheet and seed envelope. Record the cross and generation, parent identifiers, marker order and coordinates with their map or assembly, the assay call, and any ambiguous results. When a marker test is repeated, preserve both the original call and the reason it was replaced. This sounds administrative until a line with the most useful phenotype cannot be connected to the original DNA sample.
For a classroom exercise, give two people the same table of candidate genotypes and ask them independently to pick a line for a local-breakpoint study. Require each person to name the job of every column used. If one picks the highest genome score and another the shortest interval, both may be reading correctly but answering different questions. Rewrite the task until they can tell which decision has priority and what evidence could change it.
Before reporting a line as fixed for R, check zygosity. An R/r backcross plant is not the same as an R/R line. What the next generation should look like depends on the mating scheme and any effects that change transmission or survival. The marker calls that made an individual useful for a backcross are not a certificate of stability across seed generations. Plan the later checks before making the final claim.
Use a short evidence sentence: “This target-positive line had Orchard marker alleles at both named flanks; the internal marker still indicates donor ancestry on the left, and maturity is unconfirmed outside the first trial.” Notice how the sentence keeps a useful result without exaggerating it. Another reader knows exactly where to put the next assay and exactly which field claim still needs evidence.
CHAPTER 9 OF 9 · Make a claim you could defend
9. Change the context: the target is recessive and the closest marker cannot distinguish the parents
Suppose the same reasoning is needed for a recessive dry-bean seed-colour allele. A carrier may not show the colour, so rejecting every visually ordinary backcross seedling could discard the very plants needed for the next generation. Use a validated target assay or a carefully tested proxy and an appropriate confirmation strategy. In this new cross the role of the foreground assay is even clearer: phenotype in that generation cannot replace a genotype question it was never able to answer.
The closest left marker also has the same allele in both new parents. That marker cannot bound the left donor segment. A farther informative marker may let you identify promising plants, but its larger gap leaves more local ancestry unresolved. If a better marker cannot be developed, report the wider interval and use additional lines or direct phenotype evidence for decisions that cannot wait. The geometry of the first example transfers; the specific marker and screening probability do not.
One more change: the trait breeders dislike is only seen in one drying regime. The next trial should include that regime and a relevant comparison, not only the condition in which every line looks acceptable. A strong practical result answers the grower’s question in the conditions that matter. The record can distinguish three outcomes: intended allele confirmed, linked segment constrained, and product performance tested. It may reach each outcome on a different day.
Try it without the article: describe which plant you would keep if a target-positive plant has Orchard alleles at both outside markers, an unresolved inside marker, a mediocre background score and no seed-colour result yet. A good answer says what the plant is useful for now, what it cannot yet establish, and what would cause you to choose another. It does not need to call the plant either a success or a failure before the evidence is in.
Independent worked case and evidence checks
A second fictional programme has 60 target-positive seedlings. Two have a verified right recombinant, neither has a confirmed left recombinant, and three left-marker calls are unreadable. The breeder wants to print “no left recombinants in 60.” What can the record support? Start by reporting 57 readable left results with no observed left recombinant among them, plus three unresolved results. A no-call is not evidence of absence.
Next, suppose one of the two right recombinants is among those three no-calls. Do not throw away its good right-side evidence, but repeat the left assay using the tracked sample or independent material from the same identified plant. Before planning the next cross, verify the target and the right result with controls. If the left assay remains unresolved, compare the other right recombinant and decide which candidate is best for the next local test, not simply which has the prettiest summary row.
If the conditional left-recombinant probability really were 0.04 for 57 independent eligible and readable plants, the chance of no observed left recombinant would be about 0.098. That is disappointing, but possible. The calculation ceases to be a fair check if unreadable calls are related to genotype, or if the 0.04 came from a different cross. You have learned what to investigate: assay behaviour and the source of the rate, before declaring a biological barrier.
As an independent check, ask a colleague to reconstruct the candidate decision from only the marker order, parent calls, offspring identifiers and the recorded field objective. If they cannot distinguish a left no-call from a donor result, fix the record. If they can reproduce the choice but cannot tell what phenotype remains untested, fix the claim.
Worked comparison
| Observation | Supports | Does not establish |
|---|---|---|
| Target assay reads R/r | R is present under the validated assay | R is fixed or useful in every environment |
| Left and right informative markers read Orchard | The target-bearing copy has Orchard alleles at those markers when phase is resolved | An exact donor boundary or no donor DNA between markers |
| High background score | Recovery at the sampled positions under the scoring rule | A short target-linked segment |
| Maturity and resistance confirmed in a trial | Performance under stated comparison conditions | The molecular cause or performance everywhere |
Questions and practical checks
Does more backcrossing always solve linkage drag?
No. More generations offer opportunities for useful recombination. The desired allele continues to favour nearby donor material until an appropriate recombinant is identified and retained.
Can I use one flanking marker?
Yes, to examine the corresponding side if the marker is informative. Do not turn that into a two-sided claim.
Is a smaller donor segment always better?
Not necessarily. The priority depends on the breeding objective, and a neighbouring donor allele could be useful. Test the intended outcome rather than optimising segment length in isolation.
Why keep a second candidate?
An independently recovered line can provide a comparison, retain an alternative breakpoint, or preserve progress if the first line fails a different check. Label its role clearly.
Does a clean phenotype prove the donor segment is gone?
No. It shows the measured trait in the test conditions. Silent donor ancestry can remain, and a different environment may expose effects not seen in the first trial.
Are the probabilities in the article a field protocol?
No. They illustrate conditional screening under simple assumptions. Real crosses need estimated rates, suitable controls, realistic loss allowances and independent phenotype checks.
Continue with the connected subjects
Understand recombination and misleading counts
From recombination fractions to genetic map distance
Why a testcross needs a model for both inheritance and visible traits
Why some recombinant products are missing from recovered offspring
How marker checks help distinguish a stock label from a verified chromosome
Plant Breeding: how to define a useful selection goal
Return to the wider framework
Genetics: from inheritance models to independent evidence
The Gold Standard of Education: learning that transfers into life
How X Works: explore the connected subjects
Sources and scope
Frédéric Hospital, Genetics: donor chromosome segments and reducing linkage drag
Frisch and Melchinger, Genetics: the intact donor segment around a target gene
University of Hohenheim: backcross design for a recessive target allele
The numerical cases in this guide are original fictional teaching examples, not measurements from a real site. This is an educational explanation, not a field safety instruction, engineering approval or an official certification. For a real project, use qualified practitioners, current product documentation and site-specific evidence. Sources were checked on 5 November 2026.