Genomics Lab

Simplifications: a 10 kb linear genome; reads land uniformly at random with no sequencing errors (reads may hang off either end, so every base has the same expected coverage C); two reads join when they overlap by at least 20 bp. A repeat copy is resolved only when some read covers it with at least 20 bp of unique sequence on both sides; otherwise the assembly breaks there, and a greedy assembler may misjoin the flanks. Theory uses the Lander-Waterman model: uncovered fraction e^−C and expected contigs 1 + N·e^{−C(1−θ/L)} (one, plus the expected number of gaps), with θ = 20 bp.

Simplifications: equal numbers of cases and controls, an allelic test with a normal approximation, six true risk loci and about 2,200 plotted SNPs standing in for about 1,000,000 tests (the false-positive count is scaled up). SNPs near a causal locus share its signal through linkage disequilibrium. Each dot keeps its own random noise, so changing the sliders moves dots smoothly; New sample redraws the noise.

Problems are randomly generated. Numbers can be typed with or without commas, as decimals, percents (for fractions) or in e-notation (2.5e6). Answers are accepted within about 1% (fold changes and logs within 0.02). ORFs run from ATG to the next in-frame stop; count amino acids, not the stop codon. Gene trees assume no gene loss.