interpreting preseq results

[frymor]

Hi,

I am running the preseq c_curve and lc_extrap on a few

fastq

files using the option -V.
(Q: does it make sense at all to run preseq on the fastq files, or will it be more accurate, when running on the mapped files?)

I was wondering though how to interpret the results I am getting.
for example, I have a fastq file with this summarized results.
I am using the following script to run multiple files through preseq:

for file in *.fastq.gz
do
        base=$(basename $file .fastq.gz)
		zcat ${base}.fastq.gz |awk '{if (NR%4==2) print substr ($0,1,20);}'| sort | uniq -c | awk '{print $1,$2}' > ${base}.counts
		preseq c_curve  -v -V  -o ${base}.preseq.complexity ${base}.counts 2> ${base}.complexitySummary.text
		preseq lc_extrap -v -V -o ${base}.preseq.yields     ${base}.counts 2> ${base}.yieldsSummary.text
done

I get these values in the output files from the two commands (including my interpretations of the specific rows):

[B]c_curve[/B]:
VALS_INPUT
TOTAL READS     = 3130582 [I]- how many reads I have in the library[/I]
COUNTS_SUM      = 3130582 [I]- how many reads where counted in the run[/I]
DISTINCT READS  = 513863 [I]- that many distinct reads were founds[/I]
DISTINCT COUNTS = 197 [B]- what does that mean?[/B]
MAX COUNT       = 1131097 [I]- the sequence with the highest copy number[/I]
COUNTS OF 1     = 254836 [I]- number of unique reads in the library[/I]
OBSERVED COUNTS (1131098) [B]- what does that mean?[/B]

[B]lc_extrap[/B]:
VALS_INPUT
TOTAL READS     = 3130582[I] - same as above[/I]
DISTINCT READS  = 513863[I] - same as above[/I]
DISTINCT COUNTS = 197[B]- what does that mean?[/B]
MAX COUNT       = 1131097[I] - same as above[/I]
COUNTS OF 1     = 254836[I] - same as above[/I]
MAX TERMS       = 100[B]- what does that mean?[/B]
OBSERVED COUNTS (1131098)[B]- what does that mean?[/B]

the results from the two runs are as such:

[B]c_curve[/B]:
total_reads	distinct_reads
0	0
1000000	294482
2000000	414438
3000000	503167

[B]lc_extrap[/B]:
TOTAL_READS	EXPECTED_DISTINCT	LOWER_0.95CI	UPPER_0.95CI
0	0	0	0
1000000.0	294958.5	210028.8	414231.3
2000000.0	414987.0	304029.5	566439.1
3000000.0	503926.0	362806.3	699936.5
4000000.0	583253.5	410283.3	829145.8
5000000.0	658204.7	453968.8	954324.2
...
9996000000.0	8300141.2	538673.3	127892632.4
9997000000.0	8300152.7	538664.3	127895121.7
9998000000.0	8300164.1	538655.3	127897610.5
9999000000.0	8300175.6	538646.3	127900099.0

Q:Do I understand it correctly, when assuming, that in my experiment I have ~3.1M reads, from that ~255K are unique. If I'll use the same library and sequence it deeper to the depth of 9999M I will have ~8.3M unique reads?
How should I understand the two confidence intervals? (s. image below)

Q: Is there a way to say when the library is of such low quality / complexity, that it is not worth further investingating this one?
I have given here an example of what in my opinion would be not such a good library, as I have a lot of repeats (one read takes as much as a third of the data). I know there is probably no black or white in such experiments, but a rule of thumbs would be nice :-)

Q: How does the curve of the plot should look like, for a "good" and for a "bad" library?
Below are the plots I get for this library (done fast with Excel):
img=c_curve and lc_plot

thanks
Assa