Why my reads mapping rate extremely low

Question

[LeeLee]

I am using a script to analyze a large amount of ribo-seq data, which comes from different studies. Because from different research, I use trim_galore to remove the adaptor. Use bowtie2 to remove rRNA and use STAR mapping. After mapping, I found that the mapping rate of some of the data is normal, while some are extremely low, such as the following one.

UNMAPPED READS:   
Number of reads unmapped: too many mismatches  | 9953452
      % of reads unmapped: too many mismatches | 6.06%
Number of reads unmapped: too short            | 153984837
      % of reads unmapped: too short           | 93.77%
Number of reads unmapped: other                | 177470
      % of reads unmapped: other               | 0.11%

First I thought of adding parameters

--outFilterScoreMinOverLread 0 --outFilterMatchNminOverLread 0 --outFilterMatchNmin 0

, but it still didn't solve my problem.

          Time    Speed        Read     Read   Mapped   Mapped   Mapped   Mapped Unmapped Unmapped Unmapped Unmapped
                   M/hr      number   length   unique   length   MMrate    multi   multi+       MM    short    other
Oct 09 13:54:14   224.7     3931595       54     3.3%     34.2     1.1%     0.0%    96.7%     0.0%     0.0%     0.0% 
Oct 09 13:54:14   474.6    16216411       54     3.3%     34.2     1.0%     0.0%    96.7%     0.0%     0.0%    0.0%

The mapping rate after running the data with hisat2 default parameters is also similar.

Time loading forward index: 00:00:02 Time loading reference: 00:00:00
Multiseed full-index search: 00:00:04 1000000 reads; of these:  
1000000 (100.00%) were unpaired; of these:
    999572 (99.96%) aligned 0 times
     169 (0.02%) aligned exactly 1 time
     259 (0.03%) aligned >1 times
0.04% overall alignment rate

I checked the data again and confirmed that I did not mistake the species from which the data came. Why does this happen and how can I solve this problem?

[Carlo Yague]

Looks like many reads are "too short". Have you checked read length distribution with fastqc or something else ?

% of reads unmapped: too short | 93.77%

[swbarnes2]

STAR's "too short" doesn't usually mean literally too short. It just means the reads didn't map. I'd pull out the most common unmapped reads and see what they are.

[LeeLee]

Yes, I checked one of the files SRR9971635 (fastq.gz file after rRNA removal) with fastqc, and found that the length of almost all reads is 40-60bp, which is too long for ribo-seq (This is another point that confuses me).

[Jeremy Leipzig]

select one high quality sequence that you can BLAST or BLAT but doesn't align with STAR and copy-paste it here