Gene Confidence

Determined by combining BLAST agreement, reference/EST agreement, sequence quality and heuristic confidence.

1. high: Has evidence and is in clear accord with it: has strong agreement (full or partial) with reference/EST and/or strong agreement (length within +/- 25%) with BLAST evidence, no sequence problems, no model problems.
2. medium: Has evidence but is not in clear accord with it: has reference/EST and/or BLAST evidence, but either clashes with reference/EST evidence or has length within +/- 50% of BLAST evidence. Genes in which we would otherwise have high confidence, but have sequence and/or model problems, are also in this category.
3. low: Conflicts with evidence and/or has serious structural problems: has major model problems, has length that is not within +/- 50% of the BLAST homology evidence, and/or conflicts with both reference/EST evidence and BLAST evidence. Genes in which we would otherwise have medium confidence, but have sequence and/or model problems, are also in this category.
4. unknown: Has no sequence problems, no model problems, no BLAST evidence, and no reference/EST evidence.

Calculated by comparing the length of the prediction to the average length of the top 1-3 scoring overlapping BLAST hits. Each hit must have greater than 30% average identity and 30% query coverage, and the hits must come from distinct taxIds.

1. <= 10%: CDS length is within +/- 10% of the average length of the top 3 overlapping BLAST hits.
2. 11-25%: CDS length is within +/- 25% of the average length of the top 3 overlapping BLAST hits and is not within +/- 10% of that length.
3. 26-50%: CDS length is within +/- 50% of the average length of the top 3 overlapping BLAST hits and is not within +/- 25% of that length.
4. > 50%: CDS length is not within +/- 50% of the average length of the top 3 overlapping BLAST hits.
5. no evidence: Has no overlapping blast evidence of sufficient identity and coverage.

Calculated by comparing splice site predictions to those of overlapping EST clusters or reference gene models.

1. full: Has complete reference/EST coverage and no splice site conflicts.
2. partial: Has incomplete reference/EST coverage and no splice site conflicts.
3. clash: Has at least one splice site conflict with an overlapping reference/EST.
4. split: Has incomplete reference/EST coverage indicating a possible split of a single gene.
5. merge: Has incomplete reference/EST coverage indicating a possible merge of two distinct genes.
6. no evidence: Has no overlapping reference/EST evidence.

Heuristic confidence in the gene model itself, based on intron/exon lengths, splice sites, and other annotation attributes.

1. no problems: Model is longer than 90 bases long, all splice sites are canonical, no in-frame stops or frame shifts, all introns between 20 and 1000 bases (inclusive), all exons longer than 6 bases.
2. minor problems: Model is 90bp or shorter, has an intron < 20bp or > 1000bp or exon <= 6bp.
3. severe problems: Model has a frameshift, in-frame stop, is non-canonically spliced, and/or is 30bp or shorter in length.

Calculated by examining underlying and neighboring sequence.

1. no problems: No N's in exons, all basecalls higher than Q10, >1Kb from any gap(s).
2. minor problems: Within 1Kb of a sequence gap.
3. severe problems: Underlying sequence has N's within exons and/or basecalls of Q10 or lower.