Data Types

BioNodulo uses a typed port system. Every input and output port declares a data type, and the editor only allows connections between compatible types. This catches a huge class of pipeline errors before a run starts — no more feeding a BAM into a tool that expects FASTQ.

Common bioinformatics types

Type	Description	Typical producers
`FASTQ`	Raw sequencing reads (single or paired-end)	FASTQ Input, SRA Download
`FASTA`	Reference / assembled sequences	Reference Input, SPAdes
`SAM`/`BAM`	Aligned reads (text / binary)	BWA, Bowtie2, STAR
`CRAM`	Reference-compressed alignments	samtools
`VCF`/`BCF`	Variant calls	GATK, bcftools
`BED`	Genomic intervals	peak callers, annotations
`GTF`/`GFF`	Gene annotations	reference inputs
`Matrix`	Count / expression matrices (e.g. h5ad, mtx)	Cell Ranger, featureCounts
`Report`	HTML/PDF reports	FastQC, MultiQC, QUAST
`Table`	Tabular data (CSV/TSV)	many

Generic types

In addition to domain types, ports can use generic primitives:

File — an opaque file of any kind.
Directory — a folder of files.
String, Int, Float, Bool — scalar parameters and small values.
List<T> — an ordered collection of another type (used for fan-in nodes).

Coercion

Some conversions are automatic or available via explicit conversion nodes:

SAM → BAM and BAM → CRAM via samtools nodes.
BCF ↔ VCF via bcftools nodes.
A specific type can always connect to a generic File input.

The editor shows a small badge on an edge when a coercion is being applied.

Why typing matters

Earlier failure. Type errors surface at edit time, not three hours into a run.
Better suggestions. When you drag from a BAM output, the palette can suggest only the nodes that accept a BAM.
Reproducibility. Types are part of the workflow JSON, so a graph’s contract is explicit and versioned.

Defining types for custom nodes

Custom nodes declare their port types in the SDK. You can reuse the built-in types or register your own. See Creating Custom Nodes.

Nodes & Connections Execution Model