NB This method is experimental and not well tested and we invite feedback. As a result of this the two programs trace_clip and scale_trace_clip have been changed (and hopefully improved) since the last release (1996.1).
To select the clip points in the sequence we chose two measures of the trace quality.
The first (nonc, or non-called over called) calculates the ratio of the area under the trace for the called base to the maximum area under each of the non-called bases at the same position. The second (drop) measures, for the called base, the ratio of the height of the trace at its peak to its height at the mid-point between the peak and the next base. In our hands, for ABI produced traces both of these calculations produce values that start off high, drop to a minimum and then increase 5' to 3'. Example curves are shown in the two figures. In both cases the lines labelled with names ending in the letter "d" are for the "drop" calculation and those ending in "n" are for nonc. From base position 51 rightwards the calculations are averaged over window lengths of 101 bases; for base position 50 the window length is 99, for 49 it is 97, and so on, until the 5' end of the reading is reached. (Actually bases at positions 1 to MIN_LEFT are given the value at position MIN_LEFT. MIN_LEFT can be set by the user and has a default value of 5.)
Nonc and drop plots for sequence obtained using AmpliTaq. The sequence runs along the x axis and the nonc and drop values are plotted in the y direction.
Nonc and drop plots for sequence obtained using Thermo Sequenase. The sequence runs along the x axis and the nonc and drop values are plotted in the y direction.
The values plotted in these graphs were produced by the program trace_clip (see section trace_clip) whose main purpose is to clip both ends of a sequence based on the the measures drop and nonc. Trace_clip reads and analyses the trace file for a reading to produce the nonc and drop values for all positions in the sequence. Starting from the lowest point in each of the two curves (see figures) or from a user-defined base position, it then looks right until it finds the positions at which user-defined cutoff values for drop and nonc are exceeded. If stop_nonc and stop_drop are the stop positions found for the two measures then the clip point is given by:
clip = fract_nonc * stop_nonc + ( 1.0 - fract_nonc ) * stop_drop
where fract_nonc is defined by the user. That is, the clip point is defined as the weighted mean of the two stop points.