Hacker News Reader

Comment by aantix | Hacker News Reader

Nebula offers "Deep Whole Genome Sequencing - We decode 100% of your DNA at 30x"

Is the 30x coverage sufficient for finding all deleterious mutations? Or is the 100x option needed?

Coverage isn't directly relevant to whether a mutation is deleterious or not. Coverage is about getting accurate reads in the face of short-read sequencing. Modern sequencing will chop up DNA into 200-nucleotide-long fragments and then align those fragments against the reference genome. Because of the variation between individuals, sequencing errors, and the fact that there are only 4 nucleotides, one sequencing read may not be enough to unambiguously determine which DNA is present at what position in the genome. Coverage at a particular genomic coordinate or locus is the number of reads which have been aligned to that position.

Depending on the circumstances, clinical use of genomes requires 100x-500x coverage. Depending on your purposes, you may accept less sequencing depth/coverage. If I was just a hobbyist wanting some idea about my genome, 30x would be sufficient for me.

My son has been diagnosed with oppositional defiance disorder.

His symptoms greatly subside when he takes Niacin (1000mg/2x/day) along with magnesium l-threonate (2x/day).

Was wondering if there was an issue with B vitamin metabolism.

Was going to run the data through Promethease, but wasn't sure if the additional fidelity would help my search for a possible diagnosis.

I've not heard of Promethease before this discussion. It appears it will collect literature based on Single Nucleotide Polymorphisms (SNPs) in the DNA you send it.

Variant callers are a family of algorithms that look at the DNA in your sample and determine the SNPs. When thinking about coverage, quality scores assigned to sequence reads, and other pieces of information that may affect the variant caller, I would think about these inputs through their effect on detecting/classifying a variant. In other words, through metrics like accuracy, false positive rate, or false negative rate. State of the art variant callers are a little more complicated now, but simpler variant calling pipelines filter out variants that are of insufficient quality or read depth. This will primarily improve the false positive rate. If you are okay with casting a wide net in your search for a diagnosis and acknowledge that you may be seeing more false positives, then it's okay to go with a lower coverage assay.

As other commenters have pointed out, 30x coverage is widely accepted for germline sequencing. This is in contrast to somatic sequencing, which is more often done in the clinic for conditions like cancer where we expect much more variation among the DNA in the sample. Since that is not a concern in this case, you are probably fine to go for lower coverage options.

Edit: spelling

30X is considered standard for germline whole genome sequencing and higher coverage is typically used for detecting somatic variation. That being said, all detection is statistical and there are regions of the genome that are inherently hard to sequence.

30x gets you in the ballpark. there are always regions that can't be sequenced well and types of mutations that cant be reconstructed from the data. But from a clinical utility point of view a good 30x genome gets you up into the 95% range for finding the subset of variation that is clinically interpretable at the moment.

Reply on news.ycombinator.com