I have a SNP dataset aligned against the canFam4 genome assembly, and my goal is to map these variants within miRNA regions. However, all miRNA coordinates available to me were originally annotated in canFam3. Therefore, I am currently trying to convert mature miRNA coordinates from canFam3 to canFam4 using mainly LiftOver, with BLAT as an independent biological validation step.

The problem is that mature miRNAs are very short sequences, so BLAT often returns multiple perfect matches in canFam4. To solve this, I use the corresponding pre-miRNAs (~60–100 nt) as a biological anchor to identify which mature miRNA alignment is the true orthologous locus.

My workflow is the following:

1. Coordinate conversion using LiftOver I first submit the pre-miRNA coordinates from canFam3 to LiftOver in order to obtain the corresponding region in canFam4.

2. Biological validation using the pre-miRNA sequence I align the pre-miRNA sequence against canFam4 using BLAT. The rationale is that if LiftOver is correct, the actual precursor sequence should align to the same genomic region in canFam4. Since pre-miRNAs are relatively long, BLAT usually returns either a single strong alignment or very few perfect matches.

3. Interval length comparison I compare precursor sizes between canFam3 and canFam4 by subtracting start and end coordinates. The purpose is to verify whether the precursor length remains conserved and structurally intact after conversion.

4. Repeating the process for the mature miRNA I do the same for the mature miRNA sequence. However, because mature miRNAs are extremely short, BLAT returns many 100% identical hits, so BLAT alone cannot validate the LiftOver result.

5. Using the pre-miRNA as a biological filter The true mature miRNA must necessarily fall inside the coordinates of its corresponding pre-miRNA. Therefore, among all perfect BLAT hits for the mature miRNA, I consider the correct one to be the alignment located inside the pre-miRNA locus already validated through LiftOver/BLAT.

6. Validation of the mature miRNA relative position within the precursor I calculate:

   • the distance between the precursor start and mature miRNA start;
   • the distance between the mature miRNA end and precursor end;

   in both canFam3 and canFam4, and compare them. The goal is to verify whether the mature miRNA occupies the same relative position inside the precursor structure in both assemblies.

This strategy has worked well for almost all miRNAs I tested, except for one specific case: cfa-miR-8863.

In canFam3:

• pre-miRNA coordinates: chr16:13,973,805–13,973,916
• mature miRNA coordinates: chr16:13,973,830–13,973,853

Structural properties:

• precursor length: 112 nt
• mature miRNA length: 24 nt
• distance from precursor start to mature miRNA start: 25 nt
• distance from mature miRNA end to precursor end: 63 nt

After LiftOver to canFam4 using Lift Genome Annotations:

• pre-miRNA: chr16:14,657,785–14,657,902
• mature miRNA: chr16:14,657,810–14,657,833

However, structural comparison showed:

• precursor length changed from 112 nt to 117 nt
• mature miRNA length remained conserved
• distance from precursor start to mature start remained 25 nt
• but distance from mature end to precursor end changed from 63 nt to 69 nt

This suggested partial structural divergence of the precursor.

I then aligned the canFam3 pre-miRNA sequence against canFam4 using BLAT. The best result was:

• chrUn_MU018704v1:169,554–169,665

This alignment showed:

• 100% sequence identity
• complete conservation of precursor length (112 nt)
• total sequence conservation

Meanwhile, the locus obtained through Lift Genome Annotations on chr16 showed only 97.3% identity with the original canFam3 precursor sequence.

Additionally, the mature miRNA sequence aligned perfectly to 23 different loci in BLAT. Among these were:

• the LiftOver chr16 coordinate
• and a chrUn coordinate: chrUn_MU018704v1:169,579–169,602

Importantly, this chrUn mature miRNA coordinate fell exactly inside the precursor region identified by BLAT, and it perfectly preserved:

• mature miRNA length
• relative internal distances within the precursor

Another confusing observation is the following:

• when I search for cfa-miR-8863 directly in the canFam4 assembly on the UCSC Genome Browser, it points to chrUn_MU018704v1;
• but when I take the canFam3 coordinates and ask the browser to translate them into canFam4 coordinates, UCSC points to chr16.

So my question is:

Which locus should be considered the true orthologous miRNA locus in canFam4: the chr16 LiftOver result or the chrUn locus supported by perfect BLAT conservation and preserved precursor structure?

If the chrUn locus is actually the true orthologous location of this miRNA in canFam4, would it be methodologically correct to consider this miRNA as effectively “lost” for my SNP-based analyses, given that my SNP dataset does not contain variants mapped to chrUn scaffolds?

If so, how should this situation be justified methodologically in a publication or thesis? Would it be acceptable to state that the locus could not be reliably integrated into the variant dataset because the biologically supported orthologous region is located on an unplaced scaffold (chrUn) rather than on a fully assembled chromosome?