The AUMsilence platform
Target site selection and sequence design
Each sequence is designed by machine learning against the target transcript and scored against the transcriptome of the selected species. Every candidate is also checked for RNA folding, secondary and tertiary structure, protein binding, GC content and off-target matches, and its chemical modification pattern is chosen for that sequence.
Example: TP53, human
Target definition
A gene symbol and a species, human, mouse or rat, or a transcript accession where the isoform matters. For a microRNA, the mature sequence name; for a long non-coding RNA, the transcript.
- Gene
- The symbol, or a transcript accession where the isoform matters
- Species
- Human, mouse or rat, so specificity is scored against the right transcriptome
- RNA class
- mRNA, microRNA, long non-coding RNA or viral RNA, which chooses the product
- Purification and yield
- The tier and the delivered nmol, from the ladder on the order page
Transcript analysis
Every window along the target is a candidate.
Candidate scoring by machine learning
Each candidate is scored for predicted activity against the target and for specificity across the transcriptome of the selected species, so a sequence that would also bind another transcript is set aside.
The transcript, and a window on it
Reading the reference transcript.The gene's reference transcript, with a movable window along it: its position, its letters, how open the published model says it is, and how many windows of that length lie along the transcript. Sequence selection
The sequences that score best on both counts go forward. A scramble control is recommended with every order, and a labeled control shows uptake in the cells before the first knockdown experiment.
Product specification
Reading the record of the example gene.The product, the yields and grades it is sold in, and the code this design would carry, with its control. Synthesis, purification and quality control
The chosen sequences are synthesized, purified to the selected grade, checked, and shipped lyophilized. Our AUMsilence platform uses chemical modifications that enable self-delivery.
Purification and delivered yield
The grade decides how much of what arrives is full-length product, and it sets the study model.
ARPCthe example code's grade
Reversed-phase purification: the full-length oligo is separated from the shorter failure sequences by hydrophobicity. A step between desalt and HPLC, for cellular work that needs more of the tube to be full length.
in vitro, cellular studies
BHPLC
High-performance liquid chromatography resolves the full-length product from the failure sequences on a column, the highest purity of the three tiers. The grade bought for animal work.
in vivo, animal studies
Delivered yield is what arrives in the tube, in nmol, not the amount synthesized to produce it.
A lot's own measurements travel with the tube it was made for.
The purification grades this product is sold at, what each removes and the study model each is sold for, with the grade the example code carries marked. Delivery in vitro and in vivo
No transfection reagent, electroporation, or viral vector is required. Add AUMsilence
sdASOs to the culture medium, or inject them in vivo, and read the knockdown by RT-qPCR or western blot at the time the protocol gives. Published measurements
Figures from published studies that used the platform, each in the paper's own sentence. Every one is the paper's measurement, not AUM's.
What the scoring is built on
Not every position along a transcript is equally available. The RNA folds back on itself and proteins sit along it, so some sites are open to an antisense oligonucleotide and others are closed. That is why several sequences against one target do not perform alike.
The scoring rests on the published record, data reviewed from over 500,000 published articles, and the number grows every week.
That record holds the antisense and siRNA sequences that silenced their target, and the ones that did not. Each carries the cell type or the animal model it was tried in.
The same scoring runs for every sequence AUM designs, whichever product it is made as: self-delivering or transfection-optimized, antisense or siRNA. The 2'-MOE, locked nucleic acid and 2'-OMe oligonucleotides are scored the same way, wherever AUM designs the sequence rather than making one you send.
These are weighed for every candidate site along the transcript:
- the accessibility of the target site
- the secondary and tertiary structure of the RNA around it
- any protein bound there
- every off-target site the sequence could bind, found by BLAST against the species' mature mRNA, pre-mRNA, long non-coding RNA and microRNA
- the cell type the sequence is for
- the animal model, where there is one
- the GC content, and runs of a single base
- every transcript variant of the gene, so that a sequence targets all isoforms or a chosen one
- common genetic variants, SNPs, at the target site, which are avoided
- where possible, a sequence that matches human, mouse and rat alike
- the predicted binding strength with the target
- hairpins and self-pairing within the oligonucleotide, which are avoided
- the chemical modification pattern, chosen for each sequence
Why it matters
Two decisions settle an antisense experiment before it starts
Where on the transcript
Most windows on a transcript are poor targets: folded, occupied, or shared with another gene. A sequence chosen by eye against one of them fails quietly, and the failure looks like a target that cannot be silenced. Scoring every window against what the published record has shown is how the platform avoids that, before anything is synthesized.
How the sequence gets in
A sequence that never enters the cell silences nothing. Self-delivery removes the transfection step, and with it the cellular stress and the well-to-well variability the reagent brings, and it works in the primary cells conventional transfection struggles with. The design and the delivery are one product, from the same platform.
For research use only. Not for use in diagnostic or therapeutic procedures.