How to predict mutation without using variant callers like GATK, Mutect?

Question

[Mask]

I need to find out mutation in custom gene (gene of my interest). Say suppose I have 100 genes in the list I have just created a work flow. Somebody tell me how stable and sensitive it is?

• Collected nucleotide sequence of these genes and created a database of these genes using standalone blast.
• NGS analysis were carried as follows:
• Alignment of the reads ( Fastq Samples)to the reference genome)
• Created bam or sam file after alignment
• Extracted custom region (gene location of my interest) from the bam file
• Performed blast of these custom region against the database (created using standalone blast)
• In the out put, we can get variations if any:
-

Is this method correct?

[H.Hasani]

I'm missing the why?I.e. why you want to not use variants callers?

[Mask]

I need to develop a custom variant caller pipeline

Answer 1

[Petr Ponomarenko]

No, it is not correct. At best you will have erroneous SNPs predicted all over your gene with very little support (low specificity) or way too high sensitivity and will lose a lot of SNPs that can be called using proper statistic methods. What you are explaining can be done with very degraded law coverage and quality ancient DNA where there are very few good variant and haplotype callers. For a high-quality variant calling, you either need a tool dedicated for your type of samples, sequencing equipment, and chemistry, or you may try to build your own statistical model, but that makes sense nowadays only for special cases. Otherwise, you are better with using multiple tools with different parameters and aggregating data in a certain way from them, if you do not want to rely on a single tool.