Databases

and

ontologies

A machine-readable

specification

for genomics

assays

Ali Sina

Booeshaghi

1,

�

, Xi Chen

2

, Lior Pachter

3,4,

�

1

Department

of Bioengineering,

University

of California,

Berkeley,

CA, 94720,

United

States

2

Department

of Biology,

School

of Life Sciences,

Southern

University

of Science

and Technology,

Shenzhen,

518055,

China

3

Division

of Biology

and Biological

Engineering,

California

Institute

of Technology,

Pasadena,

CA, 91125,

United

States

4

Department

of Computing

and Mathematical

Sciences,

California

Institute

of Technology,

Pasadena,

CA, 91125,

United

States

�

Corresponding

authors.

Department

of Bioengineering,

University

of California,

University

Avenue

and Oxford

St, Berkeley,

CA, 94720,

United

States.

E-mail:

sinab@berkeley.edu

(A.S.B.);

Division

of Biology

and Biological

Engineering

and Department

of Computing

and Mathematical

Sciences,

California

Institute

of

Technology,

1200 E California

Blvd,

Pasadena,

CA, 91125,

United

States.

E-mail:

lpachter@caltech.edu

(L.P.)

Associate

Editor:

Christina

Kendziorski

Abstract

Motivation:

Understanding

the structure

of sequenced

fragments

from genomics

libraries

is essential

for accurate

read preprocessing.

Currently,

different

assays

and sequencing

technologies

require

custom

scripts

and programs

that do not leverage

the common

structure

of

sequence

elements

present

in genomics

libraries.

Results:

We present

seqspec

,

a machine-readable

specification

for libraries

produced

by genomics

assays

that facilitates

standardization

of

preprocessing

and enables

tracking

and comparison

of genomics

assays.

Availability

and implementation:

The specification

and associated

seqspec

command

line tool is available

at https://www.doi.org/10.5281/

zenodo.10213865.

1 Introduction

The

proliferation

of

genomics

assays

(

Ogbeide

et

al.

2022

)

has

resulted

in

a corresponding

increase

in software

for

proc

-

essing

the

data

(

Zappia

et

al.

2018

).

Frequently,

custom

scripts

must

be

created

and

tailored

to

the

specifics

of

assays,

where

developers

reimplement

solutions

for

common

prepro

-

cessing

tasks

such

as

adapter

trimming,

barcode

identifica

-

tion,

error

correction,

and

read

alignment

(

Cheow

et

al.

2016

,

Ma

et

al.

2020

,

Healey

et

al.

2022

,

Wu

et

al.

2022

).

When

software

tools

are

assay

specific,

parameter

choices

in

these

methods

can

diverge,

making

it

difficult

to

perform

apples-to-apples

comparisons

of

data

produced

by

different

assays.

Furthermore,

the

lack

of

preprocessing

standardiza

-

tion

makes

reanalysis

of

published

data

in the

context

of

new

data

challenging.

While

genomics

protocols

can

vary

greatly

from

each

other,

the

libraries

they

generate

share

many

common

ele

-

ments.

Typically,

sequenced

fragments

will

contain

one

or

several

‘technical

sequences’

such

as

barcodes

and

unique

molecular

identifiers

(UMIs),

as

well

as

biological

sequences

that

may

be

aligned

to

a genome

or

transcriptome.

Standard

library

preparation

kits

generally

require

that

DNA

from

the

libraries

is cut,

repaired,

and

ligated

to

sequencing

adapters

(Fig.

1).

Primers

bind

to

the

sequencing

adapters,

and

initiate

DNA

sequencing

whereby

reads

are

subsequently

generated.

Illumina

sequencing

employs

a sequencing

by

synthesis

ap

-

proach

where

fluorescently

labeled

nucleotides

are

incorpo

-

rated

into

single-stranded

DNA,

and

imaged,

while

PacBio

uses

zero-mode

waveguides

for

single-molecule

detection

of

dNTP

incorporation.

Oxford

Nanopore

on

the

other

hand

binds

sequencing

adapters

to

pores

in

a flow

cell

and

DNA

is

sequenced

by

changes

in

electrical

resistance

across

the

pore

(

Iizuka

et al.

2022

).

Many

single-cell

genomics

assays

introduce

additional

li

-

brary

complexity

further

complicating

preprocessing.

For

ex

-

ample,

the

inDropsv3

(

Klein

et

al.

2015

) assay

produces

variable

length

barcodes

while

the

10

×

Genomics

scRNA-seq

assay

(

Zheng

et al.

2017

) produces

fixed-length

barcodes

that

are

derived

from

a known

list

of

possibilities.

Current

file

formats

such

as

FASTQ,

Genbank,

FASTA,

and

workflow-specific

files

(

Parekh

et al.

2018

) lack

the

flexi

-

bility

to

annotate

sequenced

libraries

that

contain

these

com

-

plex

features.

In

the

absence

of

sequence

annotations,

processing

can

be

challenging,

limiting

the

reuse

of

data

that

is stored

in publicly

accessible

databases

such

as

the

Sequence

Read

Archive

(

Katz

et

al.

2022

). To

facilitate

utilization

of

genomics

data,

a database

of

assays

along

with

a description

of

their

associated

library

structures

was

assembled

in

Chen

(2020)

. While

this

database

has

proved

to

be

very

useful,

the

HTML

descriptors

are

not

machine

readable.

Moreover,

the

lack

of

a formal

specification

limits

the

utility

and

expand

-

ability

of

the

database.

2 Results

The

seqspec

specification

defines

a machine-readable

file

for

-

mat,

based

on

YAML,

that

enables

sequence

library

annota

-

tion.

Assay-

and

sequencer

specific

molecules

are

annotated

by

Regions

which

can

be

nested

and

appended

to

create

a

Received:

18

July

2023;

Revised:

4 October

2023;

Editorial

Decision:

25

January

2024;

Accepted:

4 April

2024

#

The

Author(s)

2024.

Published

by

Oxford

University

Press.

This

is an

Open

Access

article

distributed

under

the

terms

of

the

Creative

Commons

Attribution

License

(https://creativecommons.org/licenses/by/4.0/),

which

permits

unrestricted

reuse,

distribution,

and

reproduction

in any

medium,

provided

the

original

work

is properly

cited.

Bioinformatics

,

2024,

40(4)

,

btae168

https://doi.org/10.1093/bioinformatics/btae168

Advance

Access

Publication

Date:

5 April 2024

Applications

Note

seqspec

in a manner

that

assumes

perfect

end-to-end

sequenc

-

ing

of

a perfectly

constructed

library.

Regions

are

annotated

with

a variety

of

properties

that

simplify

the

downstream

identification

of

library

elements.

The

following

are

a list

of

properties

that

can

be

associated

with

a

Region

:

�

Region

ID

:

unique

identifier

for

the

Region

in the

seqspec

.

�

Region

type

:

the

type

of

region.

�

Name

:

A descriptive

name

for

the

Region.

�

Sequence

:

The

specific

nucleotide

sequence

for

the

Region.

Figure

1.

The structure

of molecules

in genomic

libraries.

Sequencing

libraries

are constructed

by combining

Atomic

Regions

to form an adapter-insert-

adapter

construct.

The

seqspec

for the assay

annotates

the construct

with

Regions

and meta

Regions

.

2

Booeshaghi

et al.

�

Sequence

type

:

The

type

of

sequence

(fixed,

onlist,

ran

-

dom,

joined).

�

Minimum

length

:

The

minimum

length

of

the

sequence

for

the

Region.

�

Maximum

length

:

The

maximum

length

of

the

sequence

for

the

Region.

�

Onlist

:

The

list

of

permissible

sequences

from

which

the

Sequence

is derived.

Importantly,

Regions

,

known

as

meta

Regions

,

can

contain

Regions;

a property

that

is useful

for

grouping

and

identifying

library

elements

that

are

sequenced

together.

The

YAML

for

-

mat

is a natural

language

to

represent

nested

meta-

Regions

in

a human-readable

fashion.

Python-style

indentation

and

syn

-

tax

can

be

used

to

create

a human-readable

file

format

without

the

excessive

grouping

delimiters

of

alternative

languages

such

as

JSON.

In

addition,

nested

Regions

allow

Assays

to

be

repre

-

sented

as

an

Ordered

Tree

where

the

ordering

of

subtrees

is

significant:

atomic

Regions

are

‘glued’

together

in

an

order

that

is concordant

with

the

design

of

the

sequencing

library

in

the

5

0

to

3

0

direction

(Supplementary

Fig.

S1).

A

key

feature

of

seqspec

files

is

that

they

are

machine-

readable,

and

Region

data

can

be

parsed,

processed,

and

extracted

with

the

seqspec

command-line

tool.

The

tool

con

-

tains

eleven

subcommands

that

enable

various

tasks

such

as

specification

checking,

finding,

formatting,

and

indexing,

1)

seqspec

check

:

check

the

correctness

of

attributes

against

the

seqspec

schema.

2)

seqspec

find

:

print

Region

metadata.

3)

seqspec

format

:

auto

populate

Region

metadata

for

meta

Regions.

4)

seqspec

index

:

extract

the

0-indexed

position

of

Regions.

5)

seqspec

info

:

get

info

about

the

seqspec

file.

6)

seqspec

init

:

initialize

a

seqspec

with

a

newick-

formatted

string.

7)

seqspec

modify

:

modify

Region

attributes.

8)

seqspec

onlist

:

get

the

path

to

the

onlist

file

for

the

spe

-

cific

region

type.

9)

seqspec

print

:

print

html,

markdown,

ascii,

read

dia

-

gram

that

visualizes

the

seqspec.

10)

seqspec

split

:

split

seqspec

into

modalities.

11)

seqspec

version

:

get

seqspec

version

and

seqspec

file

version.

To

illustrate

how

seqspec

can

be

used

to

facilitate

process

-

ing

and

analysis

of

single-cell

RNA-seq

reads,

we

imple

-

mented

in

the

seqspec

index

command

the

facility

to

produce

the

relevant

technology

string

for

three

single-cell

RNA-seq

preprocessing

tools:

kallisto

bustools

(

Melsted

et

al.

2021

),

simpleaf/alevin-fry

(

He

et

al.

2022

),

and

STARsolo

(

Kaminow

et al.

2021

) (Fig.

2).

Regions

associated

with

barc

-

odes,

UMIs,

and

cDNA

are

extracted,

positionally

indexed

and

formatted

on

a per-tool

basis.

The

modularity

of

seqspec

makes

it

simple

to

produce

tool-compatible

technology

strings

for

other

assay

types.

3 Discussion

The

seqspec

specification

and

associated

tool

enable

the

an

-

notation

of

a sequence

library

that

has

been

generated

by

an

assay

to

be

processed

with

a sequencer-specific

kit

for

se

-

quencing.

Associating

seqspecs

with

sequencing

data

can

greatly

facilitate

reprocessing

and

interpretation.

For

exam

-

ple,

seqspec

can

help

in

investigating

differences

between

se

-

quence

reads

and

library

structure,

aiding

in

the

study

of

sequencing

artifacts.

In

terms

of

facilitating

preprocessing,

seqspec

innovates

beyond

existing

methods

such

as

kb-

python’

s

technology

string

or

the

read

geometry

string

in

sim

-

pleaf

(

He

et al.

2022

) by

providing

both

annotation

of

library

structures

as

well

as

a suite

of

tools

for

format

validation.

Standardized

annotation

of

sequencing

libraries

in

a hu

-

man-

and

machine-readable

format

serves

several

purposes

including

the

enablement

of

uniform

processing,

organization

of

sequencing

assays

by

constitutive

components,

and

trans

-

parency

for

users.

The

flexibility

of

seqspec

should

allow

it to

be

used

for

all

current

sequence

census

assays

(

Wold

and

Myers

2008

), and

specifications

should

be

readily

adaptable

to

different

sequencing

platforms;

our

initial

release

of

seq

-

spec

contains

specifications

for

49

assays

(see

https://igvf.

github.io/seqspec/).

In

the

future,

we

envision

that

seqspec

could

be

extended

to

describe

sequencer-

or

protocol-specific

steps

as

well

as

utilized

to

annotate

engineered

sequences

such

as

DNA

constructs.

Comparison

of

seqspecs

for

different

assays,

immediately

reveals

shared

similarities

and

differences

that

can

be

visual

-

ized

with

seqspec

print

.

For

example,

the

SPLiT-seq

single-

cell

RNA

and

the

multimodal

SHARE-seq

single-cell

assays

are

aimed

at

different

modalities

and

utilize

different

proto

-

cols

to

produce

libraries,

but

the

resultant

structures

are

very

similar

(Fig.

1)

since

they

both

rely

on

split-pool

barcoding

(

Rosenberg

et al.

2018

). The

seqspec

for

the

sci-CAR-seq

as

-

say

(

Cao

et

al.

2018

), from

which

split-pool

assays

such

as

SHARE-seq

are

derived,

shows

that

the

cell

barcoding

is

encoded

in

the

Illumina

indices.

It

should

be

possible

to

Figure

2.

Uniform

processing

enabled

with

seqspec

.

The

seqspec

index

command

produces

a technology

string

that identifies

appropriate

sequence

elements

and can be passed

into processing

tools.

A machine-readable

specification

for genomics

assays

3

develop

an

ontology

of

assays

by

comparing

the

seqspec

spec

-

ifications

of

assays

and

quantifying

their

similarities

and

differences.

In

demonstrating

that

seqspec

can

be

used

to

define

options

for

preprocessing

tools,

we

have

shown

that

seqspec

is imme

-

diately

useful

for

uniform

processing

of

genomics

data.

The

preprocessing

applications

will

hopefully

incentivize

data

generators

to

define

and

deposit

seqspec

files

alongside

se

-

quencing

reads

in

public

archives

such

as

the

Sequence

Read

Archive.

While

seqspec

is not

a suitable

format

for

general

metadata

storage,

the

precise

specification

of

sequence

ele

-

ments

present

in

reads,

including

sequencer-specific

con

-

structs,

should

be

helpful

in

identifying

batch

effects

even

when

metadata

is missing

or

inaccurate.

Acknowledgements

We

thank

Delaney

Sullivan

for

helpful

discussions

and

Rahma

Elsiesy

for

helpful

feedback

on

Fig.

1.

Discussions

with

the

Impact

of

Genomics

Variation

on

Function

(IGVF)

Single-Cell

Focus

Group

helped

to

shape

some

features

of

seqspec

.

Thanks

to

Idan

Gabdank

for

useful

feedback

on

seq

-

spec

and

for

suggesting

the

md5

checksum.

Meichen

Fang

contributed

the

sci-RNA-seq3

seqspec.

This

work

was

pri

-

marily

undertaken

while

A.S.B.

was

at

the

California

Institute

of

Technology.

Supplementary

data

Supplementary

data

are

available

at

Bioinformatics

online.

Conflict

of interest

None

declared.

Funding

This

work

was

supported

in part

by

NIH

[5UM1HG012077-

02

to

A.S.B.

and

L.P.].

The

authors

also

acknowledge

the

Howard

Hughes

Medical

Institute

for

funding

A.S.B.

through

the

Hanna

H.

Gray

Fellows

program.

Data

availability

The

specification

and

associated

seqspec

command

line

tool

is

available

at

https://www.doi.org/10.5281/zenodo.

10213865

as

well

as

on

GitHub

https://github.com/pachter

lab/seqspec.

References

Cao

J, Cusanovich

DA,

Ramani

V

et al.

Joint

profiling

of

chromatin

ac

-

cessibility

and

gene

expression

in

thousands

of

single

cells.

Science

2018;

361

:1380–5.

Chen

X.

Collections

of

library

structure

and

sequence

of

popular

single

cell

genomic

methods.

GitHub.

2020.

https://github.com/Teichlab/

scg_lib_structs.

Cheow

LF,

Courtois

ET,

Tan

Y

et

al.

Single-cell

multimodal

profiling

reveals

cellular

epigenetic

heterogeneity.

Nat

Methods

2016;

13

:833–6.

He

D,

Zakeri

M,

Sarkar

H

et al.

Alevin-fry

unlocks

rapid,

accurate

and

memory-frugal

quantification

of

single-cell

RNA-seq

data.

Nat

Methods

2022;

19

:316–22.

Healey

HM,

Bassham

S, Cresko

WA.

Single-cell

iso-sequencing

enables

rapid

genome

annotation

for

scRNAseq

analysis.

Genetics

2022;

220

.

https://academic.oup.com/genetics/article/220/3/iyac017/652

6397.

Iizuka

R,

Yamazaki

H,

Uemura

S.

Zero-mode

waveguides

and

nanopore-based

sequencing

technologies

accelerate

single-molecule

studies.

Biophys

Physicobiol

2022;

19

:e190032.

Kaminow

B,

Yunusov

D,

Dobin

A.

Starsolo:

accurate,

fast

and

versatile

mapping/quantification

of

single-cell

and

single-nucleus

RNA-seq

data.

bioRxiv,

https://doi.org/10.1101/2021.05.05.442755,

2021,

preprint:

not

peer

reviewed.

Katz

K,

Shutov

O,

Lapoint

R

et al.

The

sequence

read

archive:

a decade

more

of explosive

growth.

Nucleic

Acids

Res

2022;

50

:D387–90.

Klein

AM,

Mazutis

L,

Akartuna

I

et

al.

Droplet

barcoding

for

single-

cell

transcriptomics

applied

to

embryonic

stem

cells.

Cell

2015;

161

:1187–201.

Ma

S,

Zhang

B,

LaFave

LM

et

al.

Chromatin

potential

identified

by

shared

single-cell

profiling

of

rna

and

chromatin.

Cell

2020;

183

:

1103–16.e20.

Melsted

P, Booeshaghi

AS,

Liu

L

et al.

Modular,

efficient

and

constant-

memory

single-cell

RNA-seq

preprocessing.

Nat

Biotechnol

2021;

39

:813–8.

Ogbeide

S, Giannese

F, Mincarelli

L

et al.

Into

the

multiverse:

advances

in single-cell

multiomic

profiling.

Trends

Genet

2022;

38

:831–43.

Parekh

S, Ziegenhain

C,

Vieth

B

et al.

zumis

– a fast

and

flexible

pipeline

to

process

RNA

sequencing

data

with

umis.

Gigascience

2018;

7

.

https://academic.oup.com/gigascience/article/7/6/giy059/5005022?

login=false.

Rosenberg

AB,

Roco

CM,

Muscat

RA

et al.

Single-cell

profiling

of

the

developing

mouse

brain

and

spinal

cord

with

split-Pool

barcoding.

Science

2018;

360

:176–82.

Wold

B,

Myers

RM.

Sequence

census

methods

for

functional

genomics.

Nat

Methods

2008;

5

:19–21.

Wu

H,

Li

X,

Jian

F

et al.

Highly

sensitive

single-cell

chromatin

accessi

-

bility

assay

and

transcriptome

coassay

with

metatac.

Proc

Natl

Acad

Sci

USA

2022;

119

:e2206450119.

Zappia

L,

Phipson

B,

Oshlack

A.

Exploring

the

single-cell

RNA-seq

analysis

landscape

with

the

scRNA-tools

database.

PLoS

Comput

Biol

2018;

14

:e1006245.

Zheng

GXY,

Terry

JM,

Belgrader

P

et

al.

Massively

parallel

digital

transcriptional

profiling

of

single

cells.

Nat

Commun

2017;

8

:14049.

#

The

Author(s)

2024.

Published

by

Oxford

University

Press.

Bioinformatics,

2024,

40,

1–4

https://doi.org/10.1093/bioinformatics/btae168

Applications

Note

4

Booeshaghi

et al.