Benchmarks / Bulk genomics & enrichment
R Bulk genomics & enrichment maftools

Speed up maftools

maftools is one of the slower steps in many bulk genomics & enrichment workflows. AutoZyme ships a verified, drop-in patch that is up to 27.0× faster, returning bit-for-bit identical results with no change to how you call it.

Best speedup 27.0×
Median speedup 23.2×
Output equivalence Bit-exact
Best runtime baseline 29.12 min optimized 1.07 min
Datasets 5
Pass rate 10/10

Benchmark charts

Switch benchmark platform; all charts update together
Platform
Speedup distribution
Each dot is one finalized dataset/thread run on Windows
log scale
laml_ood_xlargelaml_largelaml_ood_largelaml_mediumlaml_tiny
Thread sweep
Speedup across finalized thread counts on Windows
0×25×50×14full (8)laml_ood_xlarge · ood_xlarge1 threads · 27.0× speedup27.08 min baseline → 59.13 s optimizedmemory 13 GB → 9.4 GBlaml_ood_xlarge · ood_xlarge4 threads · 26.1× speedup26.11 min baseline → 1.03 min optimizedmemory 13 GB → 9.4 GBlaml_ood_xlarge · ood_xlarge8 threads · 27.0× speedup29.12 min baseline → 1.07 min optimizedmemory 13 GB → 9.4 GBlaml_large · large1 threads · 19.4× speedup12.62 min baseline → 41.94 s optimizedmemory 5.0 GB → 4.0 GBlaml_large · large4 threads · 21.4× speedup11.35 min baseline → 31.87 s optimizedmemory 5.0 GB → 4.0 GBlaml_large · large8 threads · 21.4× speedup11.64 min baseline → 33.34 s optimizedmemory 5.1 GB → 4.0 GBlaml_ood_large · ood_large1 threads · 19.5× speedup11.04 min baseline → 33.73 s optimizedmemory 5.7 GB → 3.8 GBlaml_ood_large · ood_large4 threads · 20.5× speedup9.80 min baseline → 28.35 s optimizedmemory 5.7 GB → 3.8 GBlaml_ood_large · ood_large8 threads · 18.6× speedup10.84 min baseline → 34.01 s optimizedmemory 5.7 GB → 3.8 GBlaml_medium · medium1 threads · 15.3× speedup6.27 min baseline → 24.61 s optimizedmemory 2.4 GB → 2.2 GBlaml_medium · medium4 threads · 15.4× speedup4.73 min baseline → 18.24 s optimizedmemory 2.5 GB → 2.3 GBlaml_medium · medium8 threads · 14.6× speedup5.54 min baseline → 21.98 s optimizedmemory 2.4 GB → 2.3 GBlaml_tiny · small1 threads · 9.81× speedup2.26 min baseline → 13.79 s optimizedmemory 1.2 GB → 0.8 GBlaml_tiny · small4 threads · 8.09× speedup1.53 min baseline → 11.37 s optimizedmemory 1.3 GB → 0.8 GBlaml_tiny · small8 threads · 11.3× speedup2.34 min baseline → 12.49 s optimizedmemory 1.3 GB → 0.8 GB
laml_ood_xlargelaml_largelaml_ood_largelaml_mediumlaml_tiny
Memory
Baseline vs optimized peak memory on Windows
0.0 GB10 GB20 GBlaml_ood_xlarge0.70×laml_ood_large0.66×laml_large0.79×laml_medium0.92×laml_tiny0.65×laml_ood_xlarge · ood_xlargememory 13 GB → 9.4 GBoptimized / baseline 0.70×27.0× speedup · 8 threadslaml_ood_large · ood_largememory 5.7 GB → 3.8 GBoptimized / baseline 0.66×19.5× speedup · 1 threadslaml_large · largememory 5.1 GB → 4.0 GBoptimized / baseline 0.79×21.4× speedup · 8 threadslaml_medium · mediummemory 2.5 GB → 2.3 GBoptimized / baseline 0.92×15.4× speedup · 4 threadslaml_tiny · smallmemory 1.3 GB → 0.8 GBoptimized / baseline 0.65×11.3× speedup · 8 threads
baselineoptimized

What is accelerated

The public API stays the same; AutoZyme replaces only the supported fast path.

This task targets maftools in maftools. The benchmarked result preserves the declared scientific output gate while reducing CPU runtime on the listed datasets.

Also searched as: mutation, MAF, somatic mutation, oncoplot.

Supported scope

The fast path (fast_read.maf -> fast_validateMaf + fast_summarizeMaf) handles a single MAF input that is EITHER a file path (.maf or .maf.gz, tab-separated, with a Hugo_Symbol header) OR an in-memory data.frame/data.table, when zyme=TRUE AND… Read full supported scope

The fast path (fast_read.maf -> fast_validateMaf + fast_summarizeMaf) handles a single MAF input that is EITHER a file path (.maf or .maf.gz, tab-separated, with a Hugo_Symbol header) OR an in-memory data.frame/data.table, when zyme=TRUE AND gisticAllLesionsFile is NULL AND cnTable is NULL AND isTCGA=FALSE AND useAll=TRUE. Within that envelope it correctly supports: removeDuplicatedVariants TRUE/FALSE (forwarded to validateMaf multi-key duplicated); rmFlags FALSE/TRUE/numeric (FLAG-gene removal at lines 392-402); custom vc_nonSyn (non-synonymous override at 387-391); clinicalData as NULL / data.frame / file path (handled in summarizeMaf 222-237); presence or absence of CNV variants (has_cnv branch 187-199 and Amp/Del + CNV column handling 142-170, exercised by the ood tiers); blank and NA Hugo_Symbol -> 'Unknown' (293-309); single-sample and (via nrow==0 guard) zero-variant inputs. All summary statistics (uniqueN, tabulate, colMeans/median rounded to 3, Rcpp zyme_fill_dcast integer fill) are exact equivalences to upstream; the task reports bit-exact concordance (vps_max_rel_diff=0, all *_match=1) across all 5 tiers x 2 platforms.

Out-of-scope behavior

silent fallback to upstream

Show detailed speedup table 10 runs
Dataset Tier Platform Threads Baseline Optimized Speedup Memory Concordance Pass
laml_large large Windows 8 11.64 min 33.34 s 21.4× 5.1 → 4.0 GB — pass
laml_medium medium Windows 4 4.73 min 18.24 s 15.4× 2.5 → 2.3 GB — pass
laml_ood_large ood_large Windows 4 9.80 min 28.35 s 20.5× 5.7 → 3.8 GB — pass
laml_ood_xlarge ood_xlarge Windows 8 29.12 min 1.07 min 27.0× 13.4 → 9.4 GB — pass
laml_tiny small Windows 8 2.34 min 12.49 s 11.3× 1.3 → 0.8 GB — pass
laml_large large macOS 4 7.73 min 16.38 s 28.3× 9.7 → 5.8 GB — pass
laml_medium medium macOS 8 2.80 min 7.60 s 23.6× 5.1 → 3.3 GB — pass
laml_ood_large ood_large macOS 4 5.53 min 13.55 s 24.6× 9.5 → 5.8 GB — pass
laml_ood_xlarge ood_xlarge macOS 4 15.99 min 36.88 s 26.9× 19.9 → 12.8 GB — pass
laml_tiny small macOS 8 54.37 s 2.63 s 22.7× 2.8 → 1.2 GB — pass

Frequently asked questions

Speeding up maftools
Why is maftools slow?

maftools is CPU-bound, and the stock implementation in maftools leaves performance on the table in its core numerical work. On the benchmark datasets the original takes 29.12 min where the AutoZyme path takes 1.07 min (27.0× faster).

How do I make maftools faster?

Install AutoZyme and activate the maftools patch, then keep using maftools exactly as before. AutoZyme transparently substitutes the faster, output-validated path, up to 27.0× faster on the benchmark datasets, with no pipeline or API changes.

Does the AutoZyme speedup change the maftools output?

No. The accelerated path returns bit-for-bit identical results to the original maftools implementation (maximum absolute difference 0), checked by a frozen concordance gate on every benchmark dataset.

How do I install the maftools speedup?

In R: install the autozyme package, then run library(autozyme) and autozyme::activate("maftools"). The patch applies automatically the next time you call maftools.