SeqLike - flexible biological sequence objects in Python

Introduction

A single object API that makes working with biological sequences in Python more ergonomic. It'll handle anything like a sequence.

Built around the Biopython SeqRecord class, SeqLikes abstract over the semantics of molecular biology (DNA -> RNA -> AA) and data structures (strings, Seqs, SeqRecords, numerical encodings) to allow manipulation of a biological sequence at the level which is most computationally convenient.

Code samples and examples

Build data-type agnostic functions

def f(seq: SeqLikeType, *args):
	seq = SeqLike(seq, seq_type="nt").to_seqrecord()
	# ...

Streamline conversion to/from ML friendly representations

prediction = model(aaSeqLike('MSKGEELFTG').to_onehot())
new_seq = ntSeqLike(generative_model.sample(), alphabet="-ACGTUN")

Interconvert between AA and NT forms of a sequence

Back-translation is conveniently built-in!

s_nt = ntSeqLike("ATGTCTAAAGGTGAA")
s_nt[0:3] # ATG
s_nt.aa()[0:3] # MSK, nt->aa is well defined
s_nt.aa()[0:3].nt() # ATGTCTAAA, works because SeqLike now has both reps
s_nt[:-1].aa() # TypeError, len(s_nt) not a multiple of 3

s_aa = aaSeqLike("MSKGE")
s_aa.nt() # AttributeError, aa->nt is undefined w/o codon map
s_aa = aaSeqLike(s_aa, codon_map=random_codon_map)
s_aa.nt() # now works, backtranslated to e.g. ATGTCTAAAGGTGAA
s_aa[:1].nt() # ATG, codon_map is maintained

Easily plot multiple sequence alignments

seqs = [s for s in SeqIO.parse("file.fasta", "fasta")]
df = pd.DataFrame(
    {
        "names": [s.name for s in seqs],
        "seqs": [aaSeqLike(s) for s in seqs],
    }
)
df["aligned"] = df["seqs"].seq.align()
df["aligned"].seq.plot()

Flexibly build and parse numerical sequence representations

# Assume you have a dataframe with a column of 10 SeqLikes of length 90
df["seqs"].seq.to_onehot().shape # (10, 90, 23), padded if needed

To see more in action, please check out the docs!

Getting Started

Install the library with pip or conda.

With pip

pip install seqlike

With conda

conda install -c conda-forge seqlike

Authors

• @andrewgiessel
• @maxasauruswall
• @MihirMetkar
• @ndousis
• @ericmjl

Support

• Questions about usage should be posed on Stack Overflow with the #seqlike tag.
• Bug reports and feature requests are managed using the Github issue tracker.

Contributors ✨

Thanks goes to these wonderful people (emoji key):

Nasos Dousis 💻	andrew giessel 💻	Max Wall 💻 📖	Eric Ma 💻 📖	Mihir Metkar 🤔 💻	Marcus Caron 📖	pagpires 📖
Sugato Ray 🚇 🚧	Damien Farrell 💻

This project follows the all-contributors specification. Contributions of any kind welcome!