kallisto quant 0.50.1 linux, illegal instruction

[deborah-chasman]

Hello,

I saw a few other Issues reported about "illegal instruction" behavior with kallisto index in version 0.50.
I don't think I have a problem with kallisto index (perhaps it was silent?), but I do have a problem with kallisto quant in version 0.50.1.

I was able to use an older version to complete my task successfully.

Potentially relevant info

• I'm using Scientific Linux 7.9 with kernel version 3.10.0-1160.102.1.el7.x86_64.
• I downloaded the linux binary from here:
  https://github.com/pachterlab/kallisto/releases/download/v0.50.1/kallisto_linux-v0.50.1.tar.gz

Step 1: Build an index from a custom fasta file using kallisto index

The headers from the fasta file that I supplied to the index look like this:
>sbgc_1|2051720517_214|RRE-containing [Desulfovibrio piger ATCC 29098]

Step 2: Run paired-end data through kallisto quant

I've run the same paired end data files through kallisto 0.46 and other software, so I don't expect any issues with the files themselves.

Here's the log output from kallisto quant:

[quant] fragment length distribution will be estimated from the data
Illegal instruction (core dumped)

Step 3: Attempt same with kallisto version 0.46 (success)

When I use version 0.46 (also downloaded as binary from github) to build the index on the same fasta, and to run quant on the same input, it completes with no issues.

[Yenaled]

A couple of suggestions:

1. Make sure you're using the index built with the kallisto 0.50.1 with the kallisto 0.50.1 quant (don't use an index build with an older kallistio version).
2. Try building kallisto from source:

git clone https://github.com/pachterlab/kallisto
cd kallisto
mkdir build
cd build
cmake ..
make
make install