We embedded in the astalavista framework the scoring of splice sites that are observed in the provided transcriptome annotation. By tradition in gene finding strategies [1,2,3], these are log-likelihood scores computed by the information stored in Hidden Markov Models (HMMs) which have been trained by splice sites in the corresponding species [1,4]. Pre-computed splice site models for different species can be obtained from the geneid homepage, if no explicit model is provided the astalavista scorer employs a default human HMM.
High scores represent common/well working splice sites whereas low scores reflect sites that are rare and thus probably thermodynamically also less efficient for the splicing process. By the logarithmic nature of the scores, values can drop below 0, for instance the default matrix yields values as low as (-10) for very rare splice sites. However, scores of (-9999) and lower express splice sites that have not ever been observed in the training set, which probably are non-functional.
where <annotation.gtf> is the transcriptome annotation (see GTF format) which contains the splice sites to be scored, and <genome-folder> is the path to the directory containing genomic sequences, one FASTA file per reference sequence (i.e., chromosome, contig, scaffold, etc.).
with <vcf-file> being a file in the VCF (Variant Call Format), describing polymorphisms of the genomic sequence.
where <profile> is a geneid profile for the HMM model.
For time efficiency, the positions of all genetic variants are loaded into the computer's memory (RAM), so it is to be ensured that enough memory is provided to the Java Virtual Machine. As an orientation, the variants from the 1000 Genomes project phase 1 and phase 2 just for chr22 occupy 6.4 Gb of disk and require ~1.5Gb for running splice site scoring.
Gene Models (annotation in GTF format), REQUIRED: if missing, the program responds with an error like
Chromosome sequences (FASTA files, one per chromosome), REQUIRED: if missing, the program responds with an error like
The chromosome sequences currently have to be provided as separate files, one per chromosome. All of these files have to be in the same folder (e.g., genomes/H.sapiens/hg19) with a filename prefix that corresponds to the tags in column $1 of the GTF filel provided and a suffix ".fa" or ".fasta"; e.g., if chromosomes are named "chr1", "chr2", etc. then the program expects files named "chr1.fa", "chr2.fa", ...
The first line of every
Genetic variants (as a pseudo VCF file)
Modified bases can be provided in a 5-column file format that resembles the variant-calling-format (VCF): column 1 is chromosome ID (number or letter), column 2 is position within the chromosome (integer), column 3 is variant identifier, column 4 is reference nucleotide and column 5 is the variant nucleotide.
Characters in the first column of the VCF file have to correspond to the suffixes of the chromosome names of the (1) gene annotation and (2) chromosome files, removing the prefix "chr".