Interference and Coefficient of Coincidence
Imagine mapping three linked genes on a chromosome by counting testcross progeny classes. You calculate expected double-crossover offspring by multiplying two single-crossover probabilities. This follows the ordinary rule for independent events. Then you count the actual double recombinants in your data. There are far fewer than predicted, sometimes none at all. This gap is not experimental error. It is a real biological signal. It tells you that one crossover event physically discourages a second crossover nearby. This phenomenon, called crossover interference, matters for an important reason. It reveals that recombination is not independent coin flips scattered randomly along a chromosome. Instead, the crossover machinery coordinates itself across tens of megabases. It spaces exchange points apart in a way that affects chromosome segregation. This also affects genetic map accuracy, and even fertility. Understanding interference explains why three-point mapping data almost always underestimate true genetic distances. It connects the abstract algebra of testcross ratios to the physical behavior of paired chromosomes. This physical behavior occurs in three dimensions during meiosis. The big picture is this: linkage mapping gives us a statistical map built from recombination frequencies. But the chromosome itself is a physical object. Its crossovers interact mechanically. The coefficient of coincidence (c.o.c.) is the single number that bridges these two worlds. It is simply the ratio of observed to expected double crossovers. Its departure from 1.0 quantifies interference itself. This article develops the mechanism of interference from the molecular choreography of meiotic recombination. It builds up through chromosome-scale statistics. It walks through the classical evidence that first revealed nonrandom crossover placement. It works two detailed three-point cross examples. This makes computing and interpreting the coefficient of coincidence become second nature.
- Double crossovers (DCOs) in a three-point testcross are usually rarer than the product-of-probabilities prediction — this deficit is crossover interference.
- The coefficient of coincidence = observed DCO frequency ÷ expected DCO frequency; interference = 1 − c.o.c.
- c.o.c. = 1 means no interference (independent crossovers); c.o.c. = 0 means complete interference (no DCO ever occurs in that interval).
- Interference arises because one crossover-designated site sends an inhibitory signal along the synaptonemal complex, suppressing nearby crossover formation (the "beam-film" / mechanical-stress model).
- Interference is why raw recombination frequencies underestimate true physical distance, motivating mapping functions that correct for it.
- Negative interference (c.o.c. > 1) occurs in some organisms and contexts, notably in fungi with gene conversion and in certain double-strand-break-repair pathways.
Crossover interference means one crossover physically suppresses nearby crossovers, so observed double recombinants are almost always fewer than the number predicted by treating the two crossover events as statistically independent.
The core idea
In a three-point testcross, a heterozygote for three linked genes (say A B C) is crossed to a homozygous recessive tester, and progeny are sorted into eight phenotypic classes: two parental (non-recombinant) classes, four single-crossover classes (two from each of the two intervals), and two double-crossover classes, in which a crossover occurred in both intervals simultaneously. If crossovers in the two intervals occurred completely independently of one another — as independent Bernoulli-type events, the way we assume when calculating joint probabilities of unlinked events — then the frequency of the double-crossover class should equal the product of the two single-interval recombination frequencies:
Expected DCO frequency = RF(interval 1) × RF(interval 2) This is exactly analogous to calculating the probability of flipping heads twice in a row: multiply the individual probabilities. When we actually count double-recombinant progeny in real crosses, however, the observed frequency is usually smaller — sometimes dramatically smaller — than this expectation. The gene locus order used in the calculation must first be established correctly (the middle gene is identified as the one whose allele switches partners in the double-crossover classes relative to the parental arrangement); only once the order is fixed does the interference calculation become meaningful. The coefficient of coincidence formalizes the size of this discrepancy as a single dimensionless ratio:
c.o.c. = observed DCO frequency / expected DCO frequency
Interference (I) = 1 − c.o.c. A c.o.c. of 1.0 (I = 0) means crossovers occur independently — no interference. A c.o.c. approaching 0 (I approaching 1) means complete interference: a crossover in one interval essentially guarantees no crossover in the adjacent interval. Values of c.o.c. between 0 and 1 describe positive interference, the overwhelmingly common case in eukaryotic meiosis. Occasionally c.o.c. exceeds 1.0, describing negative interference, where double crossovers are enriched relative to the independence expectation — this is rarer and typically confined to very short intervals or particular organisms and genetic contexts (e.g., gene conversion tracts in fungi).
Mechanism — how interference actually happens
Crossing over does not begin with a mature crossover; it begins with a programmed double-strand break (DSB) in DNA, catalyzed by the meiosis-specific topoisomerase-like protein Spo11, which is deliberately introduced into chromatin during early meiotic prophase I (leptotene). Hundreds of DSBs are made per meiotic cell — far more than the number of crossovers that eventually result. Each DSB is a candidate recombination site, and its resected 3′ single-stranded end invades a homologous chromatid, forming a joint molecule that can be repaired one of two ways: as a crossover (CO), involving reciprocal exchange of flanking DNA and formation of a chiasma, or as a noncrossover (NCO), in which the break is repaired without exchanging flanking markers, often via synthesis-dependent strand annealing. Only a minority of DSBs (in humans and mice, roughly 10%; in yeast, a larger fraction) mature into crossovers. Which of the many candidate DSB sites actually becomes a crossover is not left to chance alone — it is governed by an interference mechanism operating along the length of the paired homologs, which are held together at this stage by the synaptonemal complex (SC), a proteinaceous zipper-like structure that runs the length of paired chromosomes. The leading mechanistic model — the beam-film or mechanical stress-relief model, developed from cytological and physical-modeling work by Abby Dernburg, Sean Burgess, Kim Nasmyth's contemporaries, and quantitatively formalized by Andrew Zhang, Neil Kleckner, and collaborators — proposes that the chromosome axis behaves like a mechanically stressed elastic beam. When a crossover-designated recombination complex forms, it locally relieves compressive/torsional stress in the chromosome axis. This local stress relief propagates outward along the SC and suppresses the maturation of nearby DSB-repair intermediates into additional crossovers, because those nearby sites no longer experience the mechanical or biochemical trigger needed to commit to the crossover pathway. The farther away a second site is, the less the stress relief reaches it, and the weaker the interference — which is exactly why interference decays with genetic/physical distance. At the molecular level, the decision of which DSB matures into a crossover is marked by recruitment of the "ZMM" protein group (Zip1–4, Msh4-Msh5, Mer3, Spo16 in yeast, with homologs in mice/humans) to a subset of recombination intermediates; these mark the eventual class I crossovers, which are interference-sensitive. A second, smaller population of crossovers (class II, dependent on the MUS81-EME1 nuclease) shows little or no interference and contributes only a minority of total crossovers in most organisms. Because class I dominates, the pooled data from typical genetic crosses shows strong positive interference. Cytologically, the protein MLH1 forms discrete, brightly staining foci at future class I crossover sites on synaptonemal complex spreads, visible by immunofluorescence microscopy. When researchers count MLH1 foci along paired chromosomes, the foci are spaced significantly farther apart than random placement would predict — direct cytological visualization of interference, independent of any genetic cross.
The scale of interference is not uniform along a chromosome. It is strongest for nearby intervals and fades as physical distance increases — a crossover 5 map units away exerts much stronger suppression than one 40 map units away, and beyond roughly 50 cM (in many organisms), interference effects become negligible. This distance-dependence is itself part of the evidence used to build <a class="wiki-link" href="/wiki/bio-genet-genetic-map-distance-functions" data-track="wiki_article.body.wiki_link">mapping functions</a>, which correct raw recombination frequencies for interference and multiple-crossover undercounting when converting to true physical map distance.
The evidence — how interference was established
The discovery of interference did not come from imaging chromosomes; it came, decades before molecular tools existed, purely from counting flies — the same three-point testcross data used to build linkage maps, subjected to a sharper arithmetic scrutiny.
Sturtevant and Muller in the Drosophila linkage-mapping program (1910s). Once Thomas Hunt Morgan's group at Columbia began constructing multi-gene linkage maps in Drosophila melanogaster, A.H. Sturtevant noticed that when three linked genes were mapped simultaneously via three-point testcrosses, the double-crossover class was consistently rarer than the product of the two adjacent single-crossover frequencies predicted. This was not a fluke of one gene set — it recurred across chromosome after chromosome, and Hermann Muller formally named the phenomenon interference in 1916, proposing that one exchange event physically inhibited a second nearby exchange on the same chromosome pair during a single meiosis.
What exactly was measured. In a typical Drosophila three-point testcross (e.g., using the X-linked genes scute, echinus, crossveinless, or autosomal marker sets), thousands of testcross progeny were scored and sorted into the eight expected phenotypic classes. Sturtevant tabulated observed double-recombinant frequencies against the multiplication-rule expectation across dozens of independent three-point crosses. The consistent finding: observed DCO frequency was almost always less than expected, yielding c.o.c. values well below 1 in nearly every dataset, with the deficit becoming more severe as the two intervals were closer together.
What this proved. Three things were established beyond reasonable doubt by this body of data: 1. Crossovers are not statistically independent events along a chromosome — ruling out a naive model in which recombination probability at any point is set independently of what happens elsewhere. 2. The suppression effect is a function of physical/genetic proximity: intervals close together show strong interference (low c.o.c.), while widely separated intervals show interference approaching zero (c.o.c. approaching 1), consistent with a signal that propagates along the chromosome and decays with distance. 3. Interference is a chromosome-mechanical phenomenon, not a property of specific gene pairs — it was seen consistently across unrelated gene combinations and chromosomes, arguing for a general physical mechanism (later identified as SC-mediated mechanical/biochemical signaling) rather than gene-specific chemistry.
Later cytological confirmation. The genetic inference sat unconfirmed at the molecular level for most of the twentieth century until fluorescence microscopy of meiotic chromosome spreads became possible. Studies visualizing MLH1 foci (marking mature class I crossover sites) on spread synaptonemal complexes from mouse and human spermatocytes (notably work from Terry Ashley, Peter Hunt, and colleagues in the 1990s–2000s) directly showed that foci along a given bivalent are spaced significantly more evenly than a random (Poisson) distribution would produce — a direct cytological echo of what Sturtevant and Muller had inferred purely from breeding data eighty years earlier. This convergence of independent genetic and cytological evidence — one from counting phenotypes, the other from counting fluorescent dots on chromosomes under a microscope — is one of the cleanest examples in genetics of a purely statistical inference being vindicated by direct physical observation.
Worked examples
Example 1 — a Drosophila three-point testcross with asymmetric intervals. A female fly heterozygous for three X-linked genes, genotype a b c / a⁺ b⁺ c⁺ (order and phase established from parental classes), is testcrossed to a b c / Y males. 2,000 progeny are scored:
| Phenotype class | Number | Type |
|---|---|---|
| a b c | 716 | parental |
| a⁺ b⁺ c⁺ | 714 | parental |
| a b⁺ c⁺ | 82 | SCO (interval I) |
| a⁺ b c | 78 | SCO (interval I) |
| a⁺ b⁺ c | 158 | SCO (interval II) |
| a b c⁺ | 142 | SCO (interval II) |
| a b⁺ c | 6 | DCO |
| a⁺ b c⁺ | 4 | DCO |
Total = 716+714+82+78+158+142+6+4 = 1,900 (assume 1,900 fully scoreable progeny for clean arithmetic). Gene order is confirmed as a–b–c because the DCO classes (a b⁺ c and a⁺ b c⁺) differ from the parentals only at the middle locus, b — the diagnostic signature of the middle gene. Step 1 — recombination frequency, interval I (a–b): RF₁ = (82 + 78 + 6 + 4) / 1,900 = 170/1,900 = 0.0895 → map distance ≈ 8.95 cM Step 2 — recombination frequency, interval II (b–c): RF₂ = (158 + 142 + 6 + 4) / 1,900 = 310/1,900 = 0.1632 → map distance ≈ 16.3 cM Step 3 — expected DCO frequency (independence assumption): Expected = RF₁ × RF₂ = 0.0895 × 0.1632 = 0.01461 → expected DCO count = 0.01461 × 1,900 ≈ 27.8 progeny Step 4 — observed DCO frequency: Observed DCO count = 6 + 4 = 10 → observed frequency = 10/1,900 = 0.00526 Step 5 — coefficient of coincidence: c.o.c. = 10 / 27.8 = 0.36 Step 6 — interference: I = 1 − 0.36 = 0.64 (64%) Interpretation: strong positive interference. A crossover in the a–b interval suppresses a second crossover in the adjacent b–c interval roughly 64% of the time relative to what independence would predict. Note also that the true map distance (a to c) is understated by simple addition of RF₁ + RF₂ (8.95 + 16.3 = 25.25 cM) compared with what would be calculated correcting for the "hidden" double crossovers that revert to parental-looking or single-crossover-looking phenotypes — this systematic underestimate is precisely why mapping functions exist.
Example 2 — comparing interference across two genomic regions (a diagnostic/clinical-style reasoning problem). A fungal geneticist studies two different chromosomal regions in Neurospora, each spanning three linked markers, to determine whether interference strength differs between a gene-dense, low-recombination pericentromeric region and a gene-poor, high-recombination distal arm. From large-scale tetrad data: Region A (pericentromeric): RF₁ = 0.04, RF₂ = 0.03 Observed DCO frequency = 0.0009 Expected DCO = 0.04 × 0.03 = 0.0012 c.o.c. = 0.0009 / 0.0012 = 0.75 Interference = 1 − 0.75 = 0.25 (25%) Region B (distal arm): RF₁ = 0.18, RF₂ = 0.22 Observed DCO frequency = 0.030 Expected DCO = 0.18 × 0.22 = 0.0396 c.o.c. = 0.030 / 0.0396 = 0.76 Interference = 1 − 0.76 = 0.24 (24%) Surprisingly, the interference strength (c.o.c. ≈ 0.75 in both) looks similar between regions despite very different absolute recombination rates and physical densities — a reminder that c.o.c. depends primarily on genetic distance between the flanking markers, not on absolute chromosomal position or gene density per se. The geneticist correctly concludes that interference intensity should be compared for matched genetic distances, not matched physical regions: if the intervals in Region A were expanded to genetic distances equal to those in Region B, only then would a fair test of "does interference differ by chromosomal domain" be possible. This distinction — between physical position and genetic distance as the relevant variable governing interference strength — is a common source of error in interpreting real mapping data.
Common misconceptions & traps
| Misconception | Why it's wrong |
|---|---|
| "Interference means crossovers never happen twice on the same chromosome." | Double crossovers still occur — interference only reduces their frequency below the independence expectation; c.o.c. = 0 (complete interference) is a limiting case, not the norm. |
| "c.o.c. > 1 is impossible / an error in the data." | Negative interference (c.o.c. > 1) is real, though less common than positive interference; it appears in some organisms, in very short intervals, and in contexts involving gene conversion. |
| "Interference and recombination frequency are the same thing." | RF measures how often crossing over happens in one interval; interference measures whether a crossover in one interval changes the probability of a crossover in an adjacent interval. They answer different questions. |
| "You can compute c.o.c. without first establishing gene order." | Misassigning the middle gene swaps which classes count as DCOs, corrupting both RF calculations and the c.o.c. — order must be fixed first from the double-crossover class's rearranged marker. |
| "Interference is uniform along the whole chromosome." | Interference decays with distance — it is strong over short intervals and approaches zero (independence) over long intervals (often near 50 cM in many species). |
| "Interference is just a statistical artifact of small sample sizes." | It has direct cytological confirmation from MLH1-focus spacing, showing it reflects real molecular-mechanical suppression, not sampling noise. |
The scientist behind it
Interference was identified from within the extraordinarily productive Morgan "fly room" tradition at Columbia University, but its formal naming and conceptual framing is most closely credited to Hermann Joseph Muller, working alongside Alfred Sturtevant, Calvin Bridges, and Thomas Hunt Morgan as they turned Drosophila linkage data into rigorous quantitative genetics. Muller was simultaneously developing the statistical toolkit of classical transmission genetics and probing the physical nature of the gene — work that a decade later led him to discover X-ray-induced mutagenesis, for which he won the Nobel Prize in Physiology or Medicine in 1946. His recognition that double-crossover classes were systematically depleted relative to the naive independence prediction required both careful large-scale data collection (thousands of scored flies) and the willingness to trust an anomaly in the numbers as biology rather than noise — a hallmark of the quantitative rigor the fly room brought to genetics at a time when chromosomes themselves were only beginning to be accepted as the physical carriers of Mendelian factors.
Origins · 1 minHermann Joseph Muller and how this was discovered
Muller helped establish that crossovers along a chromosome are not independent events, naming the phenomenon "interference" from analysis of three-point testcross data, years before any physical or cytological explanation was available.
Where this matters today
Interference is far from a historical curiosity confined to fly-room ledgers — it has direct consequences across modern genetics.
Accurate genetic map construction. Because interference reduces the frequency of multiple crossovers, raw summed recombination frequencies (RF₁ + RF₂ + …) systematically underestimate true additive map distance over longer intervals. Mapping functions such as the Kosambi and Haldane functions explicitly build in assumptions about interference to convert observed recombination frequency into a corrected genetic distance, a concept developed further under <a class="wiki-link" href="/wiki/bio-genet-genetic-map-distance-functions" data-track="wiki_article.body.wiki_link">Genetic Map Distance and Mapping Functions</a>.
Human reproductive biology and aneuploidy risk. In human oocytes and spermatocytes, at least one crossover per chromosome arm is essentially obligatory for correct homolog segregation at meiosis I (the "obligate chiasma"). Interference helps ensure that when multiple crossovers do occur, they are properly spaced, contributing to accurate chromosome segregation; failures in crossover placement and spacing are linked to increased risk of nondisjunction and conditions such as trisomy 21, particularly in oocytes from older women, making the study of MLH1-focus spacing directly relevant to clinical reproductive genetics and IVF-related genetic counseling.
Plant and animal breeding. Breeders constructing linkage maps for crop species (maize, rice, soybean) and livestock must account for interference when estimating true genetic distances between markers flanking traits of interest, improving the precision of marker-assisted selection and QTL mapping.
Engineering recombination in biotechnology. Because interference limits how close together two crossovers can occur, genome engineers designing strategies for combining multiple linked transgenes or alleles via meiotic recombination (e.g., in synthetic biology or plant transformation stacking) must account for interference-driven suppression of nearby double exchanges when predicting recombinant frequencies.
Evolutionary and comparative genomics. Comparing interference strength (via c.o.c. across matched genetic distances) between species and even between male and female meiosis within a species (recombination rates and interference patterns differ substantially between sexes in many organisms, including humans and mice) illuminates how the meiotic machinery has been tuned by selection, informing broader questions in <span class="wiki-link-missing">Meiosis and Chromosome Segregation</span> and <span class="wiki-link-missing">Molecular Basis of Recombination</span>.
Ask the tutor: How do mapping functions like Kosambi's actually use the interference concept to convert recombination frequency into corrected genetic map distance?
Why is interference called "positive" when it reduces something (double crossovers)?
"Positive interference" refers to the sign of the effect on crossover placement: one crossover positively (actively) discourages a second nearby one, producing a real, positive inhibitory influence — not that the numeric value goes up. The confusing part is that this "positive" biological effect produces a lower observed double-crossover count and a c.o.c. below 1. "Negative interference," by contrast, describes the opposite influence — one crossover somehow makes a second nearby crossover more likely than independence predicts, producing a c.o.c. above 1. The terms describe the direction of influence between crossover events, not the direction of the resulting number.
Can the coefficient of coincidence be used to estimate physical distance between crossovers along the DNA?
Not directly on its own — c.o.c. is calculated from genetic map distances (recombination frequencies) between three marker loci, which are proxies for, but not identical to, physical (base-pair) distance. However, because interference strength correlates with physical distance along the chromosome axis (through synaptonemal-complex-mediated signaling), researchers combine c.o.c. data across many marker triplets with cytological data (like MLH1-focus spacing measured in micrometers along spread chromosomes) to build detailed models relating physical distance to interference strength, which then feed into mapping functions used to correct genetic maps.
Does interference happen the same way in mitosis?
No — the interference described here is specifically a meiotic phenomenon tied to the synaptonemal complex, programmed Spo11-induced double-strand breaks, and homologous chromosome pairing that occurs only during meiotic prophase I. Mitotic recombination is rare, occurs through different repair pathways (primarily using sister chromatids rather than homologs), and is not organized by a synaptonemal complex, so the classical crossover-interference phenomenon as measured by three-point testcrosses does not apply in the same way to mitotic recombination events.
Why do double-crossover classes matter for determining gene order in the first place?
The double-crossover progeny classes are the rarest class (because two crossovers are less likely than one), and critically, they are the only class in which the middle gene's allele appears "flipped" relative to the two flanking genes compared to the parental combination. By comparing the DCO phenotype to the parental phenotypes and noting which single gene's allele has switched partners, you identify which of the three genes lies in the middle — establishing the correct gene order before any recombination frequency or coefficient of coincidence calculation can be trusted.
Is complete interference (c.o.c. = 0) ever actually observed?
Yes, particularly over short chromosomal intervals or on shorter chromosomes/chromosome arms, where the entire region may fall within the "interference shadow" of a single crossover, making a second crossover in the same short stretch essentially impossible. In some organisms with only one or a very small number of crossovers permitted per chromosome per meiosis (a strict "one crossover per bivalent" rule, as seen in some C. elegans chromosomes), interference is effectively complete across the whole chromosome, and observed double crossovers within a single arm are vanishingly rare or absent.
How does interference relate to the "obligate crossover" needed for proper chromosome segregation?
Every bivalent (pair of homologous chromosomes) generally requires at least one crossover to form a chiasma, the physical link that ensures correct segregation at meiosis I; this is the obligate crossover. Interference works alongside this requirement by helping distribute additional crossovers (when more than one occurs per bivalent) away from each other and away from chromosome ends, promoting even spacing that likely improves the mechanical stability of chiasmata during the tension-sensing steps of meiotic segregation. Disruptions to normal interference patterning are associated with increased missegregation risk, linking this statistical genetics concept directly to chromosomal disease.
If interference lowers the observed double-crossover frequency, does it also affect the calculated single recombination frequencies (RF₁ and RF₂)?
Interference primarily affects the joint (double-crossover) frequency rather than the marginal single-interval frequencies directly, but it indirectly affects how we interpret cumulative map distance across multiple intervals. Because fewer double crossovers occur than expected, some potential "invisible" double recombination events (that would restore a parental-looking phenotype in a two-point cross) are suppressed, meaning three-point cross data with interference correction gives a more reliable estimate of true additive distance than simply adding two-point RF values calculated from separate pairwise crosses.