Select "good" reads

[a-ludi]

Selecting a "good" set of reads for the filling process is crucial to the success of the project as the most important goal is correctness.

Tasks

• [ ] Reads in a pile up should have…
  1. [x] a long “anchor” sequence, ie. presumable unique sequence
  2. [x] little overall error in the local alignments
  3. [ ] no improbable high local (per trace point) error rate
  4. [x] ~no improbable high global error rate~ (implicit with --reads-error)

Confidence intervals of local error rate

Given these values

L       ... #{reference bps covered by alignment chain}
t       ... trace point distance
n       ... either L if global or t if local
z       ... parameter for the confidence interval; multiplier for σ
ε_reads ... error rate [#{errors}/base pair] of reads
ε_ref   ... error rate [#{errors}/base pair] of reference

ε = (1 - ε_reads)(1 - ε_ref)

X_ε     ... #{errors in n bps with ε #{errors}/base pair}

we can approximate the distribution of X_ε

μ    = nε
σ²   = nε(1 - ε)
X_ε  ~ B(n, ε)
    ~= N(μ, σ²)  | valid if t >= 9/ε(1 - ε)