Malva enables RNA sequence searches across data from 74 million single cells

Malva enables RNA sequence searches across data from 74 million single cells

On August 26, Nature published a paper on Malva, an index of raw single-cell RNA reads. Malva links short sequence fragments to barcodes that identify individual cells. A query can therefore find cells containing a specific mutation, RNA junction, or evidence of a pathogen, then report the associated tissue, age, disease, and source study.

Single-cell sequencing reads the RNA from each cell separately. These reads are usually converted into a table showing how much RNA from each known gene was detected in each cell. This table helps researchers compare cells, but investigating a specific mutation, viral RNA sequence, or novel RNA junction requires returning to the large collection of raw reads.

Malva stores these reads in an index. It divides each read into 24-nucleotide fragments, in which each nucleotide is a letter in the genetic sequence, and records the cell barcode alongside them. A sequence query returns the cells containing matching fragments together with information about those cells. Indexes from separate samples can be combined, which allows new data to be added to the database.

In the version described in the paper, the index covered about 74 million cells from thousands of experiments. Searching for a 1 000-nucleotide transcript, which is an RNA copy of a gene, took 0.9 seconds on a single processor core. The search returns a pseudocount, meaning the number of matching reads rather than the number of RNA molecules.

The authors compared these pseudocounts with conventional RNA counts, then searched for sequence variants, RNA junctions, and transcript ends. In a small lung cancer dataset, Malva detected mutations in the EGFR gene. A separate analysis covered 280 samples from 16 tumor types. These tests show that rapid sequence searches can detect biological signals in cellular data.

Researchers can search existing public data for a sequence related to their hypothesis and determine the tissues, ages, and disease states in which it occurs.