The current stable release, as well as pre-compiled executable binaries for Mac OS X and GNU Linux (64bit), can be found here. The cloneHD software is undergoing rapid development. Watch/Star this repo to receive updates.

After downloading cloneHD from the release site, you can test both filterHD and cloneHD by running

$ sh run-example.sh

where you can see a typical workflow of analysing read depth and BAF data with a matched normal. All command line arguments are explained below.

For Mac OS X and GNU Linux (64bit), pre-compiled binaries are available here. To compile cloneHD yourself, you need the GNU scientific library (GSL) v1.15 or later. Change the paths in the Makefile to point to your local GSL installation (if non-standard). Then type

$ make

in the src directory. The executables will be in build. For debugging with gdb, use make -f Makefile.debug.

To report bugs, use the issue interface of github.

The full documentation can be found in the /docs/ subfolder. Click below.

• pre-filter
• filterHD
• cloneHD

cloneHD is a software for reconstructing the subclonal structure of a population from short-read sequencing data. Read depth data, B-allele count data and somatic nucleotide variant (SNV) data can be used for the inference. cloneHD can estimate the number of subclonal populations, their fractions in the sample, their individual total copy number profiles, their B-allele status and all the SNV genotypes with high resolution.

filterHD is a general purpose probabilistic filtering algorithm for one-dimensional discrete data, similar in spirit to a Kalman filter. It is a continuous state space Hidden Markov model with Poisson or Binomial emissions and a jump-diffusion propagator. It can be used for scale-free smoothing, fuzzy data segmentation and data filtering.

Visualization of the cloneHD output for the simulated data set. From top to bottom: (i) The bias corrected read depth data and the cloneHD prediction (red). (ii) The BAF (B-allele frequency), reflected at 0.5 and the cloneHD prediction (red). (iii) The total copy number posterior. (iv) The real total copy number profile. (v) The minor copy number posterior. (vi) The real minor copy number profile. (vii) The observed SNV frequencies, corrected for local ploidy, and per genotype (SNVs are assigned ramdomly according to the cloneHD SNV posterior). (All plots are created with Wolfram Mathematica.)

• The read depth input files for filterHD and cloneHD can be generated from a bam-file with samtools. Given a bed file of non-overlapping 1kb windows, e.g.

    human-genome.1kb-grid.bed :
  
    1	9000	10000
    1	10000	11000
    ...
  

  $ samtools bedcov human-genome.1kb-grid.bed sample.bam > read-depth.sample.txt

    read-depth.sample.txt :
  
    1	9000	10000	12009
    1	10000	11000	213557
    ...
  

  $ awk '{print $1,$3,int(0.5+$4/1000.0),1}' read-depth.sample.txt > read-depth.sample.cloneHD.txt

    read-depth.sample.cloneHD.txt :
  
    1 10000 12 1
    1 11000 214 1
    ...
  

  For 10 kb windows, you would use $ awk '{print $1,$3,int(0.5+$4/10000.0),10} etc. For windows smaller than 1kb, the observations might not be approximately independent.

• All input files are assumed to be sorted by chromosome and genomic coordinate. With Unix, this can be achieved with sort -k1n,1 -k2n,2 file.txt > sorted-file.txt.

• Pre-filtering of data is very important. If filterHD predicts many more jumps than you would expect, it might be necessary to pre-filter the data, removing variable regions, outliers or very short segments (use programs pre-filter and filterHD).

• Make sure that the bias field for the tumor CNA data is meaningful. If a matched normal sample was sequenced with the same pipeline, its read depth profile, as predicted by filterHD, can be used as a bias field for the tumor CNA data. Follow the logic of the example data given here.

• If the matched-normal sample was sequenced at lower coverage than the tumor sample, it might be necessary to run filterHD with a higher-than-optimal diffusion constant (set with --sigma [double]) to obtain a more faithful bias field. Otherwise, the filterHD solution is too stiff and you loose bias detail.

• filterHD can sometimes run into local optima. In this case, it might be useful to set initial values for the parameters via --jumpi [double] etc.

• By default, cloneHD runs with mass-gauging enabled. This seems wasteful, but is actually quite useful because you can see some alternative explanations during the course of the analysis.

• Don't put too much weight on the BIC criterion. It was calibrated using simulated data. For real data, it should be supplemented with common sense and biological knowledge. Use --force [int] to use a fixed number of subclones and --max-tcn [int] to set the maximum possible total copy number.

• If high copy numbers are expected only in a few chromosomes, you can increase performance by using the --max-tcn [file] option to specify per-chromosome upper limits.

• For exome sequencing data, the read depth bias can be enormous. The filterHD estimate of the bias field might not be very useful, especially in segmenting the tumor data. Use rather, if available, the jumps seen in the BAF data for both CNA and BAF data (give the BAF jumps file to both --cna-jumps and --baf-jumps).

#How to cite

The cloneHD and filterHD software is free under the GNU General Public License v3. If you use this software in your work, please cite the accompanying publication:

Andrej Fischer, Ignacio Vazquez-Garcia, Christopher J.R. Illingworth and Ville Mustonen. High-definition reconstruction of subclonal composition in cancer. Cell Reports (2014), http://dx.doi.org/10.1016/j.celrep.2014.04.055