loading...
running...
How to use
Each cycle shows one meiosis: homologs pair, chiasmata (yellow rings) form left to right and the four chromatids separate into gametes labelled by their three-locus genotype; parental classes are green, recombinants orange. The bars at the bottom and all graphs accumulate the testcross progeny of every meiosis; raise Extra meioses per second or press the +1000 buttons to collect statistics quickly. Presets set the interference strength, map length and spacing of the loci. Try: (1) No interference: coincidence goes to 1 and Haldane reproduces the map. (2) Strong interference: the double-crossover classes nearly vanish and the Poisson comparison in the last graph is much narrower than Poisson. (3) Far apart loci: r approaches 0.5, so the eight classes become equally frequent (independent assortment). (4) Tight linkage: nearly all progeny are parental. Drag to orbit, wheel to zoom. Space pauses, R restarts.
The 3D page was modelled in Blender 5.2 from named parts: a cutaway cell with nuclear envelope and spindle, the bivalent as four chromatids sliced into segments so that each segment can be coloured by the strand it comes from, locus beads, a rack of four gamete tubes, and a data screen.
The science
This is an educational Monte Carlo model of crossing over in meiosis I and of three-point testcross mapping. It is not a data-analysis tool for real crosses: real chromosomes have hotspots, sex differences and chromatid interference that the model leaves out.
Chiasmata and map distance. At pachytene the four chromatids of a bivalent exchange segments at chiasmata. Each chiasma involves one maternal and one paternal chromatid, chosen at random (no chromatid interference), so a given chromatid takes part in half of them. The genetic map unit is the Morgan: the expected number of crossovers on one chromatid. With a chiasma density of 2 per Morgan, an interval of d Morgans has on average 2d chiasmata and a single chromatid carries d crossovers. A recombinant gamete has an odd number of crossovers between two loci on its chromatid, and for N ≥ 1 chiasmata that probability is exactly 1/2, so
r = ½ ( 1 − P0(d) ), P0 = probability of no chiasma in the interval
Without interference the chiasmata form a Poisson process, P0 = e−2d and r is the Haldane function r = ½(1 − e−2d), whose inverse d = −½ ln(1 − 2r) makes map distances additive. For small d, r ≈ d (1 cM ≈ 1 % recombinants); for large d, r saturates at 1/2 (independent assortment).
Interference. Neighbouring chiasmata repel each other. The page uses the gamma model: gaps between chiasmata are Gamma-distributed with shape ν and rate 2ν per Morgan (mean spacing 0.5 M for every ν). For integer ν this process is a Poisson process of rate 2ν in which every ν-th event is kept, with the first kept event chosen uniformly among the first ν. The exact probability of an empty interval is
P0(d) = (1/ν) ∑m=1..ν ∑i<m e−2νd (2νd)i / i!
ν = 1 is no interference; large ν gives nearly evenly spaced chiasmata. The Kosambi mapping function r = ½ tanh(2d), d = ¼ ln[(1+2r)/(1−2r)], is the classic empirical compromise for moderate interference. The coefficient of coincidence c = (observed double crossovers) / (rAB rBC) measures interference as I = 1 − c: it is about 1 without interference and falls towards 0 for short adjacent intervals under strong interference. Recombination fractions are not additive (rAC < rAB + rBC because double crossovers cancel), but mapped distances are.
Numerical checks run in the page (fixed random seeds, tens of thousands of simulated meioses): (1) the mean chiasma count equals 2L; (2) and (3) the simulated recombination fractions equal the exact gamma-process formula above for one interval and for the combined interval; (4) the Haldane formula with no interference; (5) additivity of the Haldane distance recovered from rAC alone; (6) coincidence equal to 1 without interference; (7) coincidence below 0.3 for strong interference with two 15 cM intervals. The formulas are derived independently of the code that places the chiasmata.
Honest simplifications. The map length is in the chiasma coordinate, so physical (base-pair) distances, recombination hotspots, the obligate chiasma per bivalent, sex-specific maps and chromatid interference are not modelled. The gamma model needs an integer shape here. Mapping functions are idealisations: real maps are estimated from counts with sampling error, which you can see in the running estimates. Allele names A/a, B/b, C/c are generic.