discrepancy in number of reads in bam files

[baibhav-bioinfo]

Hello,

I used dorado to basecall the pod5 files into bam using m6A_DRACH model for 3 replicates of 3 conditions.
following command was used in for all replicates:
dorado basecaller [email protected] r1/pod5/ --modified-bases-models [email protected]_m6A_DRACH@v1/ --device cuda:all > r1_.bam

but when i convert the bam to fastq file, the reads in r2 and r3 are ~3 times in number and also the reads are much shorter. See the table below
Sample #reads length
Condition1_r1 10,025,268 | 1,135
Condition1_r2 33,550,922 | 180
Condition1_r3 35,376,551 | 272.1
Condition2_r1 12,537,085 | 1,107
Condition2_r2 25,225,958 | 395.5
Condition2_r3 37,304,553 | 263.5
Condition3_r1 12,715,347 | 1,065
Condition3_r2 19,535,787 | 458.4
Condition3_r3 22,782,953 | 357.9

the number of pod5 files for r2 and r3 are more in number.
can that be the reason for the difference in number of reads basecalled?

what can be the reason?

[HalfPhoton]

Hi @baibhav-bioinfo,
This is likely due to read splitting - please check out the FAQ : Why do I have more records than reads?.

It looks like your 2nd and 3rd replicate contain shorter reads or something has changed with your sample as there are many more shorter reads.

Best regards,
Rich