FAQ ...

Q: What is LiftOn?

LiftOn is a homology-based lift-over annotation tool designed to map annotations in GFF or GTF between assemblies.

LiftOn is built upon the impressive tools, Liftoff (developed by Dr. Alaina Shumate) and miniprot (Dr. Heng Li).

Q: Why do we need LiftOn?

Rather than repeatedly annotating novel genome assemblies, a more efficient strategy involves transferring genes from well-annotated organisms of the same or closely related species.

Liftoff, being entirely DNA-based, utilizes minimap2 to align gene loci DNA sequences to the genome and convert gene coordinates to the new assembly. However, when a newly assembled genome deviates significantly from the reference DNA sequence, the alignment may produce transcripts with incorrect protein-coding sequences or erroneous splice sites, posing challenges in annotation, particularly for more distantly related species.

miniprot, on the other hand, is exclusively protein-based. This approach has limitations. (1) It cannot capture untranslated regions (UTRs), (2) may miss small exons in cases of long introns, (3) is susceptible to aligning proteins to pseudogenes due to the disregard of intronic sequences, and (4) may combine coding sequences (CDSs) from distinct genes when arranged in tandem along a genome. (5) Additionally, it solely applies to protein-coding transcripts, excluding non-coding genes or other features.

To overcome these limitations, we created LiftOn, which combines the advantages of both DNA- and protein-based approaches and applies a two-step protein-maximization (PM) algorithm leading to enhanced protein-coding gene annotation.

Q: How do the annotations generated by Liftoff differ from those produced by Miniprot?

To determine how many protein-coding genes miniprot can map that are missed by Liftoff, and to assess the degree of consensus between Liftoff and miniprot regarding the coordinates of protein-coding gene loci, we investigated the results obtained by mapping the RefSeq release 220 annotations from GRCh38.p14 to T2T-CHM13 v2.

The majority of cases are '1-to-1 mapping', which includes a total of 128,351 protein-coding transcripts. Additionally, 1,986 protein-coding transcript loci are categorized under 'one-to-many mapping', with miniprot identifying a total of 7,150 transcripts. There are also 355 protein-coding transcript loci that fall under 'Liftoff-miniprot disagreement' and 334 loci under 'Liftoff misses'.

In summary, Figure A below illustrates the alignment of 130,337 (128,351 + 1,986) protein-coding transcripts with Liftoff to miniprot 1-to-1 mapping. Conversely, Figure B depicts 5,925 protein-coding transcript loci identified by miniprot as extra copies exclusive to miniprot and not overlapping with Liftoff loci."

Q: How much does LiftOn improve over Liftoff and miniprot?

Here is one example of improvement over human annotation lift-over from GRCh38 to T2T-CHM13.

Each dot represents a protein-coding transcript. If it is above the x=y line, it indicates that the LiftOn annotation possesses a higher protein sequence identity score and corresponds to a longer protein that aligns with the proteins in the reference annotation.

In the LiftOn versus Liftoff comparison (Figure above, left), 2,075 transcripts exhibit higher protein sequence identity, with 460 achieving 100% identity. Similarly, the LiftOn versus miniprot comparison (Figure above, right) discloses better matches for 30,276 protein-coding transcripts, improving 22,616 to identical status relative to the reference.

In summary, LiftOn effectively corrects quite a few protein-coding transcripts during human lift-over. The improvement is even more significant when it comes to more distant species!

Check out the Same species lift-over section, Closely related species lift-over section, and Distantly related species lift-over section for more details.

Q: How do we interpret LiftOn output?

Check out the Output files section.

Q: Can you explain the new protein-maximization (PM) algorithm in LiftOn?

Check out the Protein-maximization algorithm section.

Q: How do you evaluate the lift-over annotation?

Check out the 'Evaluation metrics - sequence identity' page.

Q: How does LiftOn report mutated genes?

LiftOn compares reference and target transcripts, similar to Liftofftools, generating a mutation report for mapped protein-coding transcripts.

Transcripts are considered "identical" if their target and reference gene DNA sequences match entirely. For mutated sequences, LiftOn categorizes changes as "synonymous", "non-synonymous", "in-frame insertion", "in-frame deletion", "frameshift", "stop codon gain", "stop codon loss", and "start codon loss".

Check out the Mutation report section.