Realign Aligned.sortedByCoord.out.bam to Aligned.toTranscriptome.out.bam for RSEM input

Question

[mkvalluru]

Hi,

I was using STAR-RSEM to quantify isoforms (primary input fastq files). Recently I have received Aligned.sortedByCoord.out.bam from the collaborators. However, I can't use these files as an input in RSEM. I have used samtools fastq to convert bam to fastq files (paired). I have used the preexisting data set for the preliminary analysis.

samtools: samtools fastq -1 Aligned.sortedByCoord.out.R1.fastq -2 Aligned.sortedByCoord.out.R2.fastq -O -t Aligned.sortedByCoord.out.bam

STAR alignment using fastq R1&R2 (samtools generated fastq files from Aligned.sortedByCoord.out.bam) generated higher unmapped reads %

        Number of input reads | 23575225
                  Average input read length |   186
               Uniquely mapped reads number |   5556353 
                    Uniquely mapped reads % |   23.57%
             % of reads unmapped: too short |   69.93%

STAR alignment using original fastq R1&R2 generated higher Uniquely mapped reads %

          Number of input reads |   23575225 
                  Average input read length |   186
               Uniquely mapped reads number |   20455386
                    Uniquely mapped reads % |   86.77%
            % of reads unmapped: too short |    7.45%

I do not know why there is a huge difference. I do not know if I am missing any steps?

-Manoj